By | Cui Peng CP
"We are just one step away from permanent capital loss."
At the end of July, when the hotshot star fund Situational Awareness (SA) sent this email to its investors, it was sliding into the abyss of collapse.
This fund, founded by former OpenAI researcher Leopold Aschenbrenner, saw its assets under management soar from $15 billion to $450 billion within a year. Just half a year ago, it was still the legend Wall Street most wanted to replicate.
Over the past few years, at various dinners in San Francisco, Aschenbrenner often talked about his dream of buying a galaxy. He believed AI was advancing so rapidly that humanity would soon be able to colonize other planets, and he needed to start saving money now.
To invest in extraterrestrial galaxies, he needed a lot of money, and SA was the best tool for making it.
The capital market also showed immense favor to this 24-year-old genius trader. Whether it was the Collison brothers, co-founders of Stripe, former GitHub CEO Nat Friedman, or quantitative giant Jane Street, all were scrambling to invest in his fund, trying to get a piece of the pie.
However, during the major correction in the semiconductor sector in July, the fund lost 67% of its investment gains in just one week, nearly triggering a chain reaction of forced liquidations. In the end, SA had to urgently sell its public market holdings to alleviate the margin crisis and avoid being forcibly liquidated by the banks.
The current narrative of AI investment has lasted nearly two years, during which various doubts failed to deter the successive waves of investors. It wasn't until SA's collapse that the market seriously began to consider a question long ignored: Does the capital market's frenzied bet on AI itself harbor systemic risk?
"The Best Opportunity to Add Positions in a Year"
Aschenbrenner's story initially seemed like another Silicon Valley and Wall Street fairytale. Public reports state that he graduated first in his class with a double major in Economics and Mathematical Statistics from Columbia University at the age of 19, then joined OpenAI's prestigious Superalignment team.
That team, co-led by OpenAI's former Chief Scientist Ilya Sutskever and Jan Leike, aimed to tackle the cutting-edge safety problem of "how to align superintelligent AI with human values."
In April 2024, OpenAI fired him for "improper disclosure of internal information." Two months later, Aschenbrenner released a 165-page report titled "Situational Awareness: The Decade Ahead," which shocked the entire industry and quickly became required reading for Silicon Valley peers.
Many investors firmly believed in his ability to judge the direction of AI technology. This reputation later directly translated into fundraising power, and the Situational Awareness hedge fund was born.
Even in the genius-filled world of Wall Street, the growth speed of this fund was staggering. According to holdings tracking data, SA started in June 2024 with $383 million in seed funding. By the first quarter of 2026, its assets under management were pushed to approximately $450 billion.
A team of fewer than twenty people, in less than two years, reached an asset management scale that many established funds operating for decades had achieved. This is extremely rare in the history of the entire fund industry.
SA's strategy was very straightforward: go long on stocks benefiting from the AI infrastructure wave (like SK Hynix, CoreWeave, and Micron), while shorting software companies deemed likely to be replaced by AI (like Adobe and AppLovin).
It sounded like a hedge structure, but in reality, it was a leveraged one-sided bet—both the long and short positions were driven by the same factor: the market's confidence in AI.
During the market upswing, this strategy brought astonishing excess returns, with the year's return reaching a peak of 439%. According to an exclusive WSJ report last week, SA's leverage efficiency was very high, reportedly able to leverage about $3 of borrowed money for every $1 of its own capital. Giants like Goldman Sachs, JPMorgan Chase, Bank of America, and Citigroup were all providing financing. Goldman Sachs even promoted the fund as a benchmark case to clients at its Emerging Managers Conference in March 2025.
It's reported that Jefferies and Barclays both rejected this business. A prime brokerage executive, after meeting with Aschenbrenner, said that his kind of unshakeable confidence was itself a danger signal. In hindsight, this made sense.
By mid-July, Wall Street began reassessing the logic of capital expenditures (AI Capex) for AI infrastructure, and the semiconductor sector officially started correcting on the 24th.
According to an FT report, Aschenbrenner sent a reassuring email to investors that day, confidently stating that this round of stock price declines was "the best opportunity to add positions in the past year," and invited investors to commit additional funds starting August 1.
Twenty Minutes from the Abyss
Now everyone knows that was not the best opportunity to add positions, but the last window to escape.
Within days, the Philadelphia Semiconductor Index fell 28.6% from its high at the end of June. SA's heavily-weighted holdings like SK Hynix, Sandisk, Bloom Energy, and Nebius all dropped over 30% in a single month.
SA's portfolio returns plummeted continuously, with the shadow of high leverage looming over the entire fund. On the morning of July 30, Goldman Sachs, JPMorgan Chase, and Bank of America simultaneously issued margin calls to SA. By then, Aschenbrenner's options were severely limited.
According to the WSJ report, the day before, SA had already urgently contacted Sequoia Capital and Greenoaks, among other Anthropic shareholders, to discuss selling private equity holdings at a discount, while also negotiating with Citadel (Citadel Investment Group). The latter's founder and CEO, Ken Griffin, joined the call from London.
The two sides finalized the deal early on the 30th, with Citadel buying all of SA's public market holdings at approximately a 10% discount. The signing was completed at 9:10 AM, just 20 minutes before the US stock market opened.
This was a true deadline: once the market opened, the banks would no longer need to follow routine procedures to gradually sell SA's holdings into the open market, avoiding a chain-reaction stampede and the disaster of more funds being liquidated.
After the transaction, SA's asset size fell from a peak of $45 billion to about $10 billion. However, public reports indicate that its holdings of Anthropic private equity were not included in this liquidation and are seen by outsiders as the foundation for SA's potential resurgence.
Some commentators eagerly painted this event as "the genius trader fell just before dawn," but this $10 billion cautionary tale should not be simply summarized as bad luck. Perhaps it is a signal that the entire AI investment narrative needs to be re-examined.
The Gamble of "Texas Hedge"
When news of SA's catastrophic losses emerged, the market's first reaction was: how could it get to this point? Didn't the fund's hedging mechanism act as a brake on the chain of losses?
Because a normal hedging strategy requires two positions to react oppositely to the same shock. Yet SA's strategy was to go long on AI beneficiaries while shorting AI victims.
The biggest problem with this strategy is that both long and short positions react in the same direction to a shock in "AI confidence." Seemingly having a short position, it actually added more leverage, multiplying the risk.
John Pfeffer, founder of family office Pfeffer Capital and an SA investor, once defended this publicly: this was never a strategy designed for conservative pension funds; everyone understood it would have剧烈 volatility.
But he and Aschenbrenner clearly underestimated the magnitude of that volatility. When the semiconductor sector declined in July, SA's long positions kept losing money, while its short positions also lost money because AI victim stocks (the software sector) rose. Getting hit from both sides resulted in SA's returns plummeting off a cliff, from a high of 439% down to 80%.
Aschenbrenner's gambler-like operations rendered the shorting mechanism virtually useless, failing to provide the intended hedge. Regarding this "pseudo-hedge" structure, Wall Street has a rather sarcastic name—"Texas hedge," akin to gambling in a game of Texas Hold'em.
Wall Street has long experienced its power. Whether it was the stock market crash triggered by the bursting of the internet bubble around 2000, or the collective collapse of quant funds in 2008, its shadow lurked behind them.
Investors, as Limited Partners (LPs), should have maintained close attention to the health of the fund's portfolio. But they often relaxed their vigilance about the fund's risk exposure because the books "had a hedge." During a one-sided market rally, both the fund and the investors rejoiced. Once the market turned downward, both would suffer losses far exceeding normal levels.
According to the WSJ report, the renowned quantitative fund Jane Street, as an investor in SA, incurred paper losses of approximately $6.5-7 billion in July alone due to SA's drawdown, astonishing the market.
Will the Market Regain Sanity Before You Go Bankrupt?
In SA's public holdings for the first quarter of this year, over 60% of its risk exposure was concentrated in a handful of AI-related stocks, layered with about 3x leverage. Furthermore, SA's holdings of private equity in Anthropic in the primary market were not quickly convertible and couldn't be used to meet intraday margin requirements, easily leading to a liquidity crisis at critical moments.
These are classic risk factors: high concentration, high leverage, and assets with medium liquidity. When they appear individually, many funds have sufficient buffers to digest the risk.
But when all three are maxed out simultaneously, it means that even a brief adverse market movement will subject the fund to multiple pressures: "losing money fast, unable to sell, and unable to meet margin calls."
Therefore, when semiconductor stocks corrected in July, it took SA less than two weeks from starting to lose money to being forced to liquidate.
Bank of America CEO Brian Moynihan said in a CNBC interview this month that this incident is a wake-up call, and mainstream prime brokers need to re-examine the risk exposures of AI-themed funds.
There are also optimistic voices in the market. The Financial Times believes that even if this round of correction is severe, SA's core judgment regarding the growth in demand for memory chips, data centers, and computing power has not yet been disproven.
The problem is, even if the long-term logic for trading the stocks holds up, the solvency pressure brought by high leverage can cause investors to fall before that logic materializes.
SA's annualized return was still around 80% at the end of July, seemingly not a small profit. But this return hinges on the fund still being alive—if Citadel's rescue hadn't arrived in time, SA would likely have faced the tragedy of being forcibly liquidated by the banks.
For a long time, there has been an investment proverb on Wall Street: Markets can remain irrational longer than you can remain solvent.
So the real question left by this crisis has long ceased to be about Aschenbrenner's judgment on AI. It is whether the capital market will continue to ignore such systemic risks to chase the next "certain" AI narrative fairy tale. (Cover image source: "Margin Call")
