Leopold Aschenbrenner launched Athena Intelligence Capital in early 2023 with no professional asset management experience. The fund now manages north of $16 billion, eclipsing Bill Ackman's Pershing Square Holdings. Ackman built his flagship over three decades. Aschenbrenner did it in 22 months.
Aschenbrenner left OpenAI in mid-2023 after publishing internal memos on AGI timelines that management found uncomfortably specific. He raised the first $500 million from family offices and sovereign wealth by September 2023, pitching a thesis that semiconductor exposure and cloud infrastructure would compress a decade of returns into 36 months. By January 2024, the fund had crossed $4 billion. Institutional allocators entered in March. The May closure brought AUM to $16.2 billion, per a source with direct knowledge of the LP base. Pershing Square Holdings reports $15.8 billion as of its last public filing.
The fund's construction is narrow. Athena holds 12 to 18 positions at any time, concentrated in GPU manufacturers, hyperscale cloud providers, and early-stage AI application companies with contracted revenue. The largest position is NVIDIA, which represents roughly 22 percent of NAV. Three cloud infrastructure names—unnamed in LP letters but identifiable through 13F filings expected in mid-February—account for another 31 percent. The remainder sits in pre-IPO stakes in companies building vertical AI agents for legal, medical, and defense workflows. Aschenbrenner does not hedge. He does not short. The fund is long-only, leveraged to 1.4x through prime brokerage lines, and rebalanced monthly based on revised AGI arrival probabilities.
What separates this from momentum chasing is the framework. Aschenbrenner's internal model assigns a 27 percent probability to AGI by 2027 and a 63 percent probability by 2030. Those figures drive position sizing. If the estimate ticks up, the fund adds leverage. If it drops, he reduces exposure but does not exit. This is not a hedge fund in the traditional sense. It is a leveraged venture portfolio dressed in liquid wrappers, run by someone who spent four years modeling transformer scaling laws at OpenAI and another two at the Centre for the Governance of AI. Allocators are paying 2 and 20 for access to a conviction set they cannot replicate internally.
The risk is obvious. If AI capital expenditure decelerates or regulatory friction emerges in Washington before late 2025, the portfolio has no cushion. The fund's December LP letter noted that a six-month delay in GPT-5 equivalent capabilities would trigger a 30 to 40 percent drawdown. Aschenbrenner has not built a hedge. He has built a directional instrument for clients who believe the next 18 months will determine the next 18 years. That is either prophecy or recklessness. The size of the bet suggests a meaningful share of the world's institutional capital believes it is the former.
Operators should watch three markers. First, whether Athena's 13F in mid-February shows increased exposure to defense AI contractors, signaling a shift toward government-backed revenue models. Second, whether the fund's leverage ratio changes in March, which would indicate a revised AGI probability estimate. Third, whether any of the pre-IPO stakes file S-1s in Q2, potentially creating liquidity events that allow Aschenbrenner to rotate capital without selling public positions. Those moves will clarify whether this is a two-year sprint or a 10-year structural reallocation.
Ackman spent 30 years building Pershing Square. Aschenbrenner spent 679 days. The difference is not skill. It is the market's willingness to pay for certainty in an uncertain domain.