Technology stocks have bounced. After a turbulent stretch for AI-related names in July, the tech-heavy NASDAQ Composite is up more than 8% from its lowest levels last month and is now reapproaching record highs.
Strong earnings and an improved outlook for a peace in the Middle East have been among the key drivers—but another, largely invisible development appears to have contributed to both the sell-off and the recovery.
Over the course of July, Situational Awareness—the extraordinarily successful AI-focused hedge fund founded by 24-year-old Leopold Aschenbrenner—became trapped in a vicious cycle of falling prices, margin calls and forced selling.
The fund had accumulated large, leveraged positions across the AI infrastructure trade. As those stocks declined, it was forced to unwind most of a public equity portfolio that had reached as much as $16 billion.
These positions were ultimately sold in large part to Citadel, the investment firm founded by Ken Griffin—the seasoned fund manager who has been a major player across the capital markets for decades.
Leopold’s pain appears to be Ken’s gain.
It has been reported that Citadel’s multi-billion flagship fund gained nearly 6% in July, after having taken over the Situational Awareness public equity portfolio at a 10% discount.
When the novice money manager got overextended and urgently needed liquidity, the wily veteran was standing by, ready to lend a helping hand… for a price.
Leverage and volatility claim another victim
Just two years ago, Leopold was asked on a podcast about his top priorities for the investment firm he was preparing to launch.
“Obviously, not blowing up is task number one and two,” he replied.
Last week, it nearly did.
It is impossible to know precisely how much of the July decline in AI stocks was caused by the problems at Situational Awareness—or how much of the recovery reflects the completion of its forced liquidation.
But the episode, like the scores of other Wall Street fiascos that have preceded it, is worth studying, as it offers investors several important takeaways.
These lessons that extend well beyond one hedge fund—or even the recent turmoil in AI stocks.
It shows how hidden market forces can drive stock prices far from underlying fundamentals, and how forced selling can create volatility that investors mistakenly interpret as new information about a business.
It demonstrates how extreme leverage can transform a painful but survivable correction into an existential crisis, stripping an investor of the ability to wait for a long-term thesis to play out.
It also reveals Wall Street’s recurring tendency to confuse intelligence, confidence and a spectacular early track record with durable investment skill—even in the presence of obvious red flags (like a complete lack of investing experience).
We spoke to one prominent asset allocator this week, who informed us that his investment committee had in fact “scolded” him earlier in the year for failing to invest with Leopold—showing just how easily even the most allegedly sophisticated investors can get carried away by groupthink and exciting personal narratives.
Above all, the Situational Awareness wipeout is a reminder that successful investing is not simply about identifying where the world is going.
Successful investing is about building a portfolio that is capable of surviving unexpected turns along the way.
Leopold’s fund has apparently not been fully liquidated. It still owns private market investments, including an early stake in AI model platform Anthropic, a business which could be on its way to becoming one of the most valuable businesses in the world.
Nonetheless, this was a catastrophic mishap for investors in the Situational Awareness fund, which declined in value by nearly 70% in July. And for the general investing public, it was a source of needless market volatility.
On the other hand, it creates a good learning opportunity.
Below, we distill six important investment lessons from the events of the past few weeks. But first…
Who is Leopold Aschenbrenner?
To understand how the fund came so close to total disaster, it helps to understand the unusual young man who created it.
Leopold does not have the conventional pedigree of a hedge fund manager.
Ken Griffin famously traded bonds and other securities from his Harvard dorm room (after persuading the school to permit a rooftop satellite dish that supplied real-time market data). He later made his mark in the intensely quantitative world of convertible bond arbitrage.
Leopold had a much different backstory. Rather than learning the ropes as an analyst or trader, he got into money management through his early close associations with the budding AI community.
Born and raised in Germany, he graduated from Columbia University in 2021 as valedictorian at just 19, majoring in economics and statistics.
After college, Leopold worked at the FTX Future Fund, the philanthropic arm of Sam Bankman-Fried’s crypto empire, where he advocated for a political philosophy known as Effective Altruism (EA).
Proponents of the EA movement claim they seek to use evidence and reason to determine how money and talent can be used to generate the maximum amount of good, including by protecting future generations from potential threats such as advanced AI.
Critics of EA view the movement as highly technocratic, elitist and arrogant—placing extraordinary confidence in a small group of self-appointed, wealthy, highly educated people (as opposed to democratically elected leaders) to calculate which causes, risks and even human lives deserve the most attention.
While there is no suggestion that Leopold participated in the fraud that later brought down FTX, the job placed him inside an elite network of young intellectuals, technology entrepreneurs and wealthy donors.
At the Future Fund, he met Avital Balwit, who is now his wife and chief of staff to Anthropic CEO Dario Amodei. (Remarkably, the young couple in fact got married just last weekend at a lavish ceremony in California—in the immediate aftermath of the blow-up.)
Leopold later joined OpenAI (provider of ChatGPT) and studied how increasingly powerful AI systems could be controlled. But OpenAI actually fired him in 2024 over an alleged leak of internal information, a characterization he disputes.
A very powerful idea
Leopold’s close proximity to the emerging AI world gave him early insight into its massive potential.
Shortly after leaving OpenAI, he published Situational Awareness: The Decade Ahead. In this 165-page essay, he argued that Artificial General Intelligence (AGI)—the moment when machines becomes just as intelligent as people—could arrive much sooner than most people expect.
The core thesis of the essay rested on compounding improvements in computing power, algorithms and the “unhobbling” of AI models through better tools. Together, he argued, those forces could advance AI capabilities much faster than conventional forecasts assumed.
This insight had important investment implications. If he was right, the world would need extraordinary quantities of chips, memory, data centers and electricity.
With this extremely bullish point of view on AI growth, Aschenbrenner turned the essay into a hedge fund bearing the same name.
He initially raised a few hundred million dollars from prominent Silicon Valley figures, including Stripe founders Patrick and John Collison, former GitHub CEO Nat Friedman and investor Daniel Gross.
Despite his youth and inexperience as a fund manager, the prestige of those early backers supplied credibility. Spectacular returns, driven by early positioning in some of the biggest AI winners as well as financial leverage, supplied the rest.
By June 2026, Situational Awareness reportedly managed more than $20 billion. It had gained approximately 270% after fees through May of this year and more than 1,000% since inception.
Even Jane Street—one of the world’s most sophisticated trading firms and an unusual investor in outside managers—had committed capital.
Wall Street had found its newest genius.
Then July happened. The value of the fund’s portfolio fell 67% in just one month.
Situational Awareness had to sell most of its publicly traded holdings, eliminated its leverage entirely and retreated largely into private investments, including its extremely valuable Anthropic stake.
Information about the Situational Awareness debacle continues to trickle out. Many more details will likely be revealed in time.
But beyond the entertainment value of the drama, like most investment disasters, it already offers some very valuable practical lessons.
Ken Griffin may have been the immediate financial beneficiary, but the rest of us can get something out of this as well….
(1) Price action is not always information
Markets are constantly producing prices. They are not always explaining why those prices are moving.
When an AI stock falls sharply, commentators immediately construct a fundamental narrative….
The AI boom is ending. Hyperscaler spending has become irrational. Memory prices have peaked. Chinese competition has permanently damaged the industry.
Some of these concerns are legitimate.
But a lot of the pressure on AI stocks in July was clearly linked to this particular hedge fund’s troubles. The stocks had to be sold because of what was happening to the fund’s balance sheet—not because of what was happening inside the businesses.
Price action can contain valuable information, but it can also reflect factors that have nothing to do with long-term value.
Never automatically assume a price decline means the prospects of a business have declined. It just means incremental selling pressure exceeds buying pressure at that moment.
(2) Leverage turns volatility into risk of ruin
Without leverage, Situational Awareness would have experienced an extremely painful month. With leverage, it experienced permanent capital impairment.
Leverage does more than magnify gains and losses. It changes who controls the portfolio.
When an investor owns stocks with cash, a 30% decline is unpleasant—but the investor can choose to wait.
When those stocks are financed with borrowed money, declining prices reduce the collateral protecting the lender. The lender can demand more capital, reduce financing or liquidate positions.
The process becomes an automatic one-way ratchet. Falling prices trigger margin calls. Margin calls require selling. Selling pushes prices lower. Lower prices produce more margin calls.
At that point, the manager is no longer making investment decisions. The manager is negotiating with the banks.
With three or four times leverage, which Situational Awareness was reportedly using, the sort of decline an unleveraged investor might survive can wipe out most of a fund’s equity.
Leopold may ultimately be totally correct about AI and its vast potential. But leverage eliminated his ability to wait for the future he predicted.
(3) Intelligence, performance and experience are not the same thing
Leopold is clearly a “high IQ individual,” as the expression goes.
Graduating from Columbia as valedictorian at 19, working on frontier AI research and producing an influential technological forecast are serious accomplishments.
But none of these prove that someone knows how to manage institutional money at scale.
The contrast with Ken Griffin is instructive. Griffin founded Citadel in 1990 and spent more than three decades building an institution designed to survive market stress.
When Situational Awareness needed to sell, Citadel had the capital, systems and experience to buy.
Leopold possessed the foresight. Griffin possessed the staying power. In the final stage of the trade, staying power was more valuable.
(4) Conviction can become dangerous concentration
Aschenbrenner’s unusual conviction was initially his edge. He saw the potential scale of the AI infrastructure buildout before most investors understood it and was willing to make large bets.
The early results appeared to validate both the thesis and the intensity with which he expressed it.
But conviction can be a feature that turns into a bug.
A portfolio may own semiconductors, memory manufacturers, data centers, power producers and AI cloud providers.
On a spreadsheet, those look like different investments across several industries. Economically, they may still represent one trade.
All depend on accelerating AI capital spending, available financing, investor enthusiasm and high valuation multiples. When those assumptions come under pressure, apparent diversification disappears.
Situational Awareness did not necessarily own too few securities. It owned too many versions of the same idea.
Rapid asset growth makes the problem worse. A strategy managing several hundred million dollars can move relatively easily. A fund overseeing tens of billions has fewer places to invest and much less room to exit.
Scale does not just make the bets larger. It makes the exits smaller.
(5) A great thesis is not a complete investment process
The fund’s wipeout does not prove Leopold’s AI thesis is wrong. It does not prove hyperscalers will fail to earn acceptable returns or that demand for memory, computing power and electricity will disappoint.
Many of the underlying companies could still produce excellent results. But a technological forecast is not an investment process.
A complete process must answer different questions: What price should be paid? How large should the position become? How much of the portfolio depends on the same assumptions? Can the investor survive being early?
Aschenbrenner developed a sophisticated theory of where technology was heading. He did not construct a portfolio capable of surviving every path the market might take to get there.
(6) Survival is the ultimate investment advantage
Risk management is often seen as a constraint—a cautious discipline that prevents investors from fully capitalizing on their best ideas.
In reality, it is what allows compounding to continue.
Investors do not need to capture 100% of the upside from every transformational trend. They need enough exposure to benefit if they are right without taking so much risk that an ordinary reversal permanently removes them from the market.
Situational Awareness’s recent investor letter said the fund came “unacceptably close” to permanent capital impairment. It has now removed all leverage and retained valuable private investments, so Leopold may indeed fight another day.
He may also be proved right about the long-term direction of AI. But that does not erase the lesson.
No technological forecast—however intelligent, detailed or ultimately correct—can eliminate the oldest risks in investing.
Prices fluctuate. Popular trades reverse. Liquidity disappears precisely when investors need it most.
The goal is not to build the portfolio that produces the greatest possible return if everything goes according to plan. It is to build one that can continue compounding when something inevitably does not.
