Age 24, a former OpenAI researcher, was elevated to god-like status in Silicon Valley by a 165-page manifesto, and built a $45 billion hedge fund in two years, with returns once exceeding 1000%.
However, this fund nearly collapsed within a month. The day the crash finally materialized coincided with the arrival of guests for his wedding.
The story recently added another chapter: The SEC has issued subpoenas to Goldman Sachs, JPMorgan Chase, Citigroup, and Bank of America, demanding communications records related to the fund's trading timing and leverage usage. It should be noted that this inquiry is still in its early stages, and the fund has not been charged with any misconduct.
This is not just a story about a young man's leveraged blow-up. How did a young man with no investment background manage to raise $45 billion in two years, bypassing Wall Street's entire due diligence system? Who was truly behind this collapse?
The Making of a Legend: How a Manifesto Translated into $45 Billion
Aschenbrenner's resume itself reads like a high-potential script.
At 19, he graduated as the top student with a double major in Economics and Mathematical Statistics from Columbia College, Columbia University. In 2023, he joined OpenAI's "Superalignment" team, considered one of the company's core groups possessing the most advanced judgment on AGI safety.
In 2024, he was fired by OpenAI. The official reason was that he leaked sensitive information to external parties; he argued that he was merely discussing hypothetical scenarios about AI safety with outsiders.
This controversy did not ruin his reputation; instead, it created a tragic narrative of a "prophet ostracized for telling the truth." In Silicon Valley's context, this archetype of the "visionary exiled by a big corporation" has inherent virality.
After being fired, he wrote a 165-page long essay titled "Situational Awareness." Its core thesis was: The composite curves of computing power expansion, algorithmic efficiency improvement, and model capability growth make the emergence of AGI as early as 2027 "surprisingly plausible," and it will trigger an "intelligence explosion" affecting national security and the geopolitical landscape. (Note: Original article link: https://situational-awareness.ai/ )
This article quickly became known as "must-read in Silicon Valley," with its influence, according to multiple media reports, even reaching policy discussions in Washington.
From then on, he was no longer an obscure 24-year-old but was labeled as "one of the few who truly sees the future trajectory of AI."
Moreover, he lived in the same house as an Anthropic senior employee, renowned AI chip supply chain analyst Dylan Patel, and podcast host Dwarkesh Patel. This circle itself is one of the most well-informed microcosms in the AI industry.
The label was soon cashed in for real money.
The fund was backed with capital from Stripe co-founders Patrick Collison and John Collison, as well as investors Daniel Gross and Nat Friedman.
Early bets also initially validated the narrative: The fund successfully bet on AI infrastructure beneficiaries like SK Hynix, CoreWeave, and Bloom Energy, achieving a 47% net return (after management and performance fees) in the first half of 2025.
Notably, an independent market analyst mentioned in a personal blog post that this fund, in its first year of operation with only $225 million under management and a founder with zero investment track record, managed to secure an allocation in CoreWeave's IPO. Such an opportunity is almost impossible to obtain under traditional asset management's mature due diligence processes.
A "Perfect Hedge" That Became a Double Whammy
The fund's prospectus for Situational Awareness stated: "no restrictions" on investment types, and no limits on "investment concentration or the amount of leverage used."
This was not an execution error; it was by design—there were no brakes installed in the system.
Its trading structure appeared, on the surface, to be a classic long-short hedge: Go long on the physical bottlenecks of AI infrastructure, including memory chips, computing power, and energy; simultaneously short traditional software companies deemed to be disrupted by AI. This structure was, for the past eighteen months, a "beautiful trade."
The problem was that these two legs were actually different expressions of the same macro factor, not truly opposite-direction hedges.
In July of this year, both bets reversed simultaneously: the market grew skeptical about the sustainability of AI infrastructure spending, with heavy holdings like SK Hynix and CoreWeave falling over 30% in a month; meanwhile, shorted traditional software stocks like Adobe began to rebound over 20%.
The long leg was falling, and the short leg, which should have been profitable, was also losing money. A true hedge requires opposite directions under the same factor; what occurred here was a "rising and falling together." The higher the leverage, the more fatal this two-way pummeling becomes.
And the leverage was pushed to the extreme. According to estimates from multiple media outlets, the fund's leverage ratio was as high as 400%, meaning for every dollar of its own capital, it borrowed four dollars to amplify bets.
At such leverage levels, a 35% to 47% drop in core holdings is enough to wipe out the equity capital itself.
One market analyst's comment was: "Running a 400% lever with such a concentrated position is not a strategy, it's a countdown."
The high profile itself also became a risk amplifier in this collapse. When the fund reduced positions, the massive and highly concentrated sell orders in a few AI-related stocks made it easy for other traders to guess who the seller was, further exacerbating the stampede.
Ultimately, the fund sold its leveraged public equity holdings at a discount to Citadel.
A Recurring Wall Street Script
This is not the first time Wall Street has paid the price for a "sufficiently compelling narrative."
In 2021, Bill Hwang's family office, Archegos Capital Management, using similar tactics—concentrated holdings, high leverage, and hidden true exposure through total return swaps—ultimately triggered combined losses exceeding $10 billion for banks including Credit Suisse, which was severely weakened as a result.
This time, a new actor appeared in the script: Jane Street.
This notoriously secretive market-making giant, known for its strict quantitative risk controls, was a major investor in Situational Awareness. According to reports from Reuters, CNBC, Fortune, and others, Jane Street's exposure to this fund reportedly snowballed from an initial ~$2.5 billion to nearly $10 billion, before falling back to around $3 billion after the July crisis.
This loss directly led to Jane Street's first overall monthly loss in nearly a decade, forcing it, led by JPMorgan Chase, to urgently issue $14.6 billion in new bonds to refinance approximately $11 billion in existing floating-rate loans and debt.
How could a man in his twenties, who had never managed money before, borrow such massive amounts sufficient to lever up 400%?
According to industry media reports, one family office invested in this fund because its owner, himself the founder of a large US tech company, personally instructed the investment, thereby bypassing his own institution's originally established risk control procedures.
This seems more like a structural tendency rather than a precise, proven causal chain.
The bankers approving the financing have their bonuses and review cycles calculated quarterly and annually, while the risk of a leveraged blow-up only materializes when the market truly reverses—a classic principal-agent problem. The decision-makers enjoy the immediate gains; the institutions and shareholders bear the future risks.
The banks also don't need to judge whether the long-term AI infrastructure trend is right; they only need to judge "Can I detect the risk earlier than the fund itself, tighten conditions earlier, and offload the risk earlier?"
Meanwhile, in Q2 2026, Goldman Sachs' equity financing business revenue hit a record $3.26 billion, 26% higher than its entire 2020 revenue, with market demand "far exceeding the amount banks were willing to take on." When all peers are fighting for the same pie, a single bank tightening standards on its own means watching revenue literally walk into competitors' hands.
Another real cognitive bias is also at play: When a fund selectively discloses returns that once hit 1000%, many investors subconsciously take this stellar performance as evidence of the founder's "foresight." Carrying this bias, those involved ignore many clues pointing to "this might just be a combination of directional bets layered with high leverage that hasn't exploded yet."
An Isolated Case or a Trend?
Looking at this single event, the banks didn't lose money, the fund wasn't liquidated, there was no systemic market shock, and Citadel even booked paper gains after taking over the positions at a discount. From this angle, it resembles a stress test more than a real crisis.
But if we zoom out and look at the total leverage within the entire AI sector?
The Bank for International Settlements (BIS) issued a warning in 2026: The combined AI capital expenditures planned by leading cloud providers for 2025-2026 could exceed $1 trillion. A reversal in this investment wave could potentially drag some economies into recession.
Bank of England Deputy Governor Sarah Breeden also voiced a related concern from another angle: If automated trading systems, trained on similar data and optimized for similar objectives, make synchronized selling decisions under stress, it could trigger a self-reinforcing crash faster than human traders could react.
These warnings are not idle talk. According to FINRA data, US securities margin debt reached $1.416 trillion in May 2026, a record high since the data series began in 1997, representing a staggering 54% year-over-year increase. According to Goldman Sachs citing EPFR statistics (this data only covers 573 US leveraged equity ETFs, excluding bond, commodity leveraged products, or overseas markets like Korea), the total assets of such US leveraged equity ETFs doubled from $39 billion in April this year to $84 billion by the end of May. JPMorgan Chase CEO Jamie Dimon also publicly acknowledged that current margin debt levels are "quite high."
Intriguingly, merely days after the crisis, Aschenbrenner secured another $400 million in private investment for a new bet, bringing his total personal wager to $500 million, including a previous $100 million.
He clearly does not view this near-collapse as the end of his investing career.
The greatest irony this saga leaves behind might be this: A person who once specialized in researching "how to anticipate and manage the risk of AGI going rogue" personally designed a leverage structure for his own hedge fund that was most prone to running out of control.
Understanding risk and managing risk are two entirely different capabilities. Seeing the direction of a grand trend does not equate to having the ability to navigate its short-term volatility unscathed.
This crisis didn't destroy any institution, nor did it truly interrupt this young man's own investment rhythm. But it tore open a fissure, allowing a glimpse into just how fragile the mechanisms meant to stop risk have become behind the AI narrative-driven capital frenzy.
The next time a narrative is compelling enough, will there be someone willing to sound the alarm first? (This article was first published on Titanium Media APP, author | Silicon Valley Tech_news, editor | Lin Shen)
