Leopold Aschenbrenner, the former OpenAI safety researcher who published "Situational Awareness" in June 2024, filed his first 13F quarterly disclosure for a hedge fund bearing the same name. The filing shows options positions in NVIDIA and adjacent AI semiconductor names totaling north of $100 million notional exposure. Aschenbrenner left OpenAI in April 2024 after internal disputes over AGI timelines and compute governance.
The 13F lists call options on NVIDIA as the dominant position, with smaller allocations to TSMC, Broadcom, and AMD. The structure suggests directional conviction rather than volatility arbitrage—long-dated calls purchased during the September semiconductor selloff when NVIDIA traded near $102 before its recent recovery to $138. The fund launched in Q3 2024 with capital from undisclosed backers, timing its entry after NVIDIA's post-earnings drawdown and before the October rally that added $600 billion to the company's market capitalization.
Aschenbrenner's transition from policy writing to capital deployment marks a clean example of the researcher-to-allocator pipeline that emerged after the GPT-4 deployment cycle. His June essay argued that AGI arrival by 2027 would hinge on U.S.-China compute competition and domestic semiconductor capacity—a thesis that now has seven-figure backing. The options structure allows levered exposure to NVIDIA's earnings volatility without requiring the full capital base of a long-only equity fund, a setup that makes sense for a first-time manager raising from family offices and crypto-native allocators who remember the 2021 ARK Innovation ETF playbook.
The disclosure matters because it confirms that former frontier-lab researchers are now running institutional capital against their own timelines. Aschenbrenner is not buying NVIDIA for the data center cycle or the Blackwell ramp—he is buying it because his internal models say compute is the bottleneck to AGI, and NVIDIA is the only name with the margin structure to survive a three-year capex war between Anthropic, Google DeepMind, and whatever OpenAI becomes. The options expiry dates, if disclosed in future filings, will telegraph whether he expects the next compute milestone in 2025 or 2027.
Operators should track NVIDIA's January 2025 earnings call for any forward guidance on H200 allocation, since Aschenbrenner's thesis depends on supply tightness persisting through 2026. Watch for follow-on 13F filings in April 2025 to see whether he rolls the options forward or converts to equity. If other OpenAI or Anthropic alumni file similar disclosures, it signals that the researcher class believes the AGI race is a public-market trade, not just a venture-stage bet.
The filing landed the same week NVIDIA reclaimed its $3.4 trillion market cap, making Aschenbrenner's entry price look disciplined rather than heroic. The real tell will be whether he adds to the position if NVIDIA pulls back to $120 on profit-taking, or whether the options were a one-time translation of a PDF thesis into a Bloomberg position.