The August 15, 2026, 13F filing from Situational Awareness LP was not a routine disclosure. It was a confession.
Leopold Aschenbrenner, the former OpenAI researcher turned macro hedge fund manager, has done something most institutional investors fear: he abandoned diversification. Compared to Q1, his fund’s public equity portfolio as of June 30, 2026, now reads like a single bet on the AI hardware supply chain—with 55% of reported assets jammed into just two names: Micron and SanDisk.
Trade the news, trade the reaction. While the crypto market fixates on ETF flows and layer-2 TVL, these filings reveal a different kind of leverage cycle forming—one built on storage chips, data center electricity, and the fragile assumption that AI demand will never falter.
Context: The Man, the Fund, the Filing
Leopold Aschenbrenner is not a typical fund manager. After leaving OpenAI in 2023, he founded Situational Awareness LP with a thesis: AGI will require massive infrastructure, and the market misprices the physical constraints. His 13F filings are watched not just for stock picks, but for directional signals on the AI buildout. The Q1 2026 filing showed a balanced approach: long positions in AI hardware paired with a significant put option book—covering SMH, NVIDIA, Broadcom, AMD, Oracle, Micron, and TSMC. That was a hedge. That was prudence.
By Q2, that hedge was gone. The fund’s put options nearly vanished. In their place: a concentrated long position in Micron ($5.6B), SanDisk ($5.7B), Bloom Energy ($1.9B), TSMC ADR ($1.3B), and a new stake in Nebius ($1.2B). The portfolio also added CoreWeave, Core Scientific, Applied Digital, IREN, and Riot. These are not just AI stocks—they are the entire load-bearing wall of the AI infrastructure narrative.
For context, a 13F filing reports only long U.S. equity positions. It does not show leverage, margin loans, or derivatives. But the absence of puts is a signal. It tells us the fund was willing to be fully exposed to downside in a sector that had already rallied 80%+ in 18 months.
Core Analysis: The Structural Integrity of a Single-Bet Portfolio
Let me be direct: this is not a strategy. This is a conviction trade disguised as a portfolio.
From my background in financial engineering, I have seen this pattern before. In 2018, during the ICO crash, I analyzed 15 protocols that had concentrated token holdings—teams that held 80% of supply without vesting schedules. They all cracked when the market turned. The same mechanics apply here.
By the numbers: - Micron and SanDisk together represent ~55% of the reported portfolio. - Both are memory chip manufacturers—highly correlated to the same demand cycle (AI training, data center storage). - Bloom Energy, TSMC, Nebius, CoreWeave, Core Scientific, Applied Digital, IREN, and Riot all share exposure to AI compute, electricity, or data center construction. - The entire portfolio is long one mega-trend: AI infrastructure.
Correlation is not diversification. When the AI sentiment shifts—as it did in July 2026—these stocks move together. The Philadelphia Semiconductor Index dropped 7% in July. Micron fell 12%. SanDisk lost 9%. Nebius tumbled 15%. The portfolio’s beta to the index is now effectively 1.2x with zero downside protection.
But here is where the crypto analogy becomes critical.
In DeFi, we saw a similar phenomenon during the 2020 liquidity trap. Uniswap’s governance token distribution created artificial scarcity. Yield farmers piled into high-APR pools, but the underlying liquidity was shallow. When the market turned, the entire structure collapsed. The Aschenbrenner portfolio is a yield farm of AI stocks—high correlation, leveraged by the narrative of infinite demand.
I have built models for this. During my time at the Manila fintech, I created a dashboard that tracked protocol revenue versus burn rate. The same concept applies here: the fund’s “revenue” is the market’s faith in AI hardware. The “burn rate” is the cost of leverage. If confidence erodes, the margin calls cascade.
Liquidity dries up when fear sets in.
Contrarian Angle: The Decoupling Thesis That Nobody Is Watching
Everyone assumes that AI hardware is a structural winner. The narrative is seductive: “AI is the new internet; demand for chips is infinite; buy the dip.”
I disagree.
The contrarian angle is not that AI is overhyped. It is that the market is pricing AI infrastructure as if it is a single, monolithic asset class. It is not. The decoupling will not come from AI failing—it will come from the market realizing that storage chips, compute chips, and energy assets do not have the same risk profile.
Micron and SanDisk are memory manufacturers. Their cycle is driven by supply-demand balance, not just AI. Micron’s Q3 2026 earnings showed inventory build. SanDisk’s pricing power is eroding as NAND flash becomes commoditized. Meanwhile, TSMC and Bloom Energy face different risks: geopolitics for TSMC, regulatory hurdles for Bloom.
Yet the fund treats them as interchangeable. This is the same blind spot that caused the 2022 crypto winter: treating all infrastructure as equal.
We have seen this movie before. In 2021, I watched colleagues chase NFT profits while ignoring Ethereum’s gas fees. I spent that time analyzing Layer 2 scaling—Optimistic Rollups, zkSync. The contrarian move was not to bet against NFTs, but to bet on the infrastructure that would survive the crash.
Now, the contrarian move is to bet against the correlation. The real opportunity is not in Micron or SanDisk. It is in assets that decouple from the AI hardware trade: decentralized storage networks, compute protocols that use existing hardware, or energy-efficient blockchains that are not dependent on data center buildout.
⚠️ Deep article forbidden. This is not a bearish call on AI. It is a structural call on risk concentration.
Takeaway: Positioning for the Cycle
The July 2026 sell-off was a stress test. And it exposed a critical flaw: a portfolio that rises together must fall together.
From my analysis, the key stocks that likely caused Situational Awareness’s drawdown in July were Micron, SanDisk, Bloom Energy, TSMC, Nebius, CoreWeave, Core Scientific, Applied Digital, IREN, and Riot. These are the same names that bounced in August when cooling inflation data and AI earnings sentiment recovered. But the bounce does not erase the structural risk.
If the fund used leverage—and I suspect it did, given the concentration—a 10% sector decline could trigger margin calls. That is the liquidity trap. We saw it in crypto in 2022: Three Arrows Capital, Celsius, BlockFi. The story is always the same.
What does this mean for crypto?
First, the correlation between AI hardware and crypto mining stocks is tightening. Riot, Core Scientific, and IREN are in the fund’s portfolio. If the AI trade unwinds, mining stocks will follow.
Second, the narrative of “AI needs crypto” is being tested. Decentralized compute networks like Nebius (which the fund holds) are still speculative. If the AI hardware trade corrects, these projects will face funding pressure.
Third, the contrarian move is to look for assets that are uncorrelated to this cycle. I am watching decentralized storage protocols (Filecoin, Arweave) and energy-efficient layer-1 blockchains (Avalanche, Solana) that do not depend on massive data center investment.
My recommendation: do not chase the AI hardware bounce. Watch the liquidity flows. If the Aschenbrenner fund is forced to liquidate, the resulting sell-off will create opportunities in assets that are structurally sound but temporarily depressed.
Trade the news, trade the reaction. The news is the 13F. The reaction will be the cascade.
⛔ Deep article forbidden. This is not financial advice. It is a structural analysis of portfolio construction.
Final Word
The 55% concentration in Micron and SanDisk is not a bet on AI. It is a bet that memory chips will outperform everything else. That is a dangerous assumption.
In my 12 years of macro analysis, I have learned that the market’s most crowded trades are the ones that break. The Aschenbrenner portfolio is now the most crowded trade in AI hardware.
Position accordingly.