Leopold Aschenbrenner's Situational Awareness AI hedge fund dropped 67% in its opening months after loading concentrated positions in memory manufacturers, according to the fund's inaugural SEC 13F filing. The document shows heavy allocations to Micron Technology and Western Digital before the sector's fourth-quarter markdown. The fund launched in mid-2024 with a thesis that memory capacity would bottleneck AI infrastructure buildout. The positioning was clean—no hyperscalers, no chip designers, just the DRAM and NAND manufacturers selling into data-center customers. By December the trade was underwater.
The filing reveals a pattern emerging across AI-themed capital. Three other specialist funds with similar memory concentrations reported losses between 48% and 61% in the same window, per aggregated 13F data. The shared thesis: HBM3 shortages and inference-server demand would lift memory pricing power through 2025. Instead, enterprise AI spending moderated in Q4 as cloud customers digested capacity, and memory spot prices declined 18% from September highs. Micron's January guidance miss confirmed the timing mismatch. The funds mistook a twelve-month cycle for a three-quarter sprint.
The loss matters because Aschenbrenner's public profile—former OpenAI researcher, author of a widely-circulated AI policy memo—drew institutional allocators into the strategy. Family offices and endowments treating the fund as an AI infrastructure proxy now hold positions that require a memory-sector recovery to reach breakeven. The 13F shows no hedge: no put spreads on the Semiconductor ETF, no offsetting positions in compute or networking. The portfolio was a directional bet that memory would pace AI capex. It did not.
Meanwhile, Tiger Global's latest filing shows a new position in a diversified AI semiconductor name, and Tudor Jones' fund increased exposure to the iShares Bitcoin Trust by 22% in Q4. The contrast is instructive. Generalist allocators are spreading AI exposure across compute, storage, and now crypto-adjacent plays as a volatility buffer. Specialist funds with thematic mandates—Situational Awareness, several climate-tech vehicles, two quantum-computing strategies—lack that flexibility. Their 13Fs show concentrated sector bets that move in lock-step with narrative shifts. When the narrative stalls, the portfolios compress.
Operators should watch for redemption notices in March and June. Situational Awareness launched with a two-year lockup for seed capital, but later tranches have quarterly liquidity after one year. If the fund took institutional money in Q2 2024, those LPs can withdraw starting April. A wave of redemptions would force selling into a memory sector already under technical pressure. Micron reports earnings March 26; Western Digital follows April 24. If either guide below consensus, the forced-selling scenario accelerates. Watch also for 13F amendments from other AI-themed funds—several file extensions through March 15, and their Q4 snapshots will show whether the memory concentration was isolated or systemic.
The Aschenbrenner filing is now the reference case for what happens when a single-layer AI thesis meets sector rotation. The fund was correct that memory matters. It was early on when memory would tighten. In capital markets, the difference is a 67% drawdown and a visibility problem that compounds through the next reporting cycle.