After a few years working in Silicon Valley, with an essay titled “Situational Awareness” predicting the AI-future, Leopold Aschenbrenner took the tech-world by storm, becoming an overnight sensation. Then, propelled by some of the world’s deepest pockets, he set his sights on the financial world.

His next story was arguably even more sensational in scale and popularity, but, again, the subject matter was not unheard of. After all, people racing to call the end of the world and AI-bulls are not in short supply – in fact, they often seem to be one and the same.

Whether you believe that the end of humanity as we know it, at the hands of AI, is a bull case, or not, it seems some do; as late this past July a hedge-fund called Situational Awareness – Aschenbrenner’s latest escapade by that same name – shot to the top of the ‘biggest trading loss of all time’ leaderboard.1

Biggest trading losses of all time leaderboard (July 2026 USD). Source: Financial Times Alphaville / Wikipedia.

The financial strategy behind losing $35 bn

In a recent article covering this fund’s place in finance’s ‘hall of fame’, Toby Nangle at the FT wrote a two-step guide for any other fund that is interested in “losing a lot of money”. Summarising the recipe, the steps are rather simple: convince wealthy people to give you money; then do away with conventional financial wisdom and throw “a bunch of financial leverage into the mix for good measure”.1

Given your intention, really, is to lose a large sum of money, the first part – convincing people to give you money – is the hardest to execute. Rest assured, however, if you follow the strategy employed by Situational Awareness, even if the assets you choose tend upwards, losing money is almost a mathematical certainty.

Indeed, this fund did pick good stocks, with the book gaining 439% in the first 6 months of 2026.2 So, for half a year, their heavily leveraged bet on an AI bull run seemed like a stroke of genius. Unfortunately, it only took a third of that time for the maths, and then the margin calls, to come knocking.

The stocks the fund picked did not crash to historic lows and there was no bursting of the alleged AI-bubble that haunts Wall Street – in fact the S&P 500 remains around all-time-highs. There was only an AI - sentiment wobble that caused a large fall in the recent months’ hottest stocks – like SK Hynix – and a historic momentum churn; with the sector-neutral Momentum Index curated by Morgan Stanley posting its worst decline ever.3

Nevertheless, due to its strategy, the fund didn’t need a full stock-market disaster to topple.

Positive expected return; negative real return?

The reason why the fund took off was the same reason it crashed: leverage. Aggressively borrowing money to pile into AI stocks allowed the fund to make spectacular returns, but it also caused a week’s selloff to create catastrophic losses. To understand why the fund’s thesis was right but they still ended up selling almost everything, at a discount, to Ken Griffin’s Citadel, we must look at a concept known as volatility drag.

Volatility drag is the phenomenon that explains the gap between the expected return (arithmetic average) and the true return a portfolio experiences (the compound annual growth rate, or geometric mean). It stems from the percentage asymmetry in investments and presents the uncomfortable reality that volatility alone can erode returns. You can pick the right assets; have the right thesis; guess the right long-term direction; calculate a positive average return; and volatility alone can put you in the red.

It works on the basis that losses reduce the base capital that is supposed to compound. If a portfolio rises 5% then falls 5%, the average return is 0%, but the compound return is -0.25% – the loss is applied to a larger pie and thus symmetrical moves can leave you below break even. The important result that follows is that the magnitude of this effect is a function of the size of the swings and thus is a function of volatility (a 20% rise followed by a 20% fall also results in an average return of 0%, but a compound return of -4%).

That is the lesson that the fund learnt the hard way: when leverage is introduced, the maths starts to care more about volatility than direction. With this concept in mind, we can then see why leverage is a double-edged sword, given its sole purpose is to turn 5% ebbs and flows into 20% rises and falls.

The approximate relationship between the expected (arithmetic) return, ; the true (geometric) return, ; and volatility, , is: 4

Herein lies the reason that leverage is not a magic wand: your expected return grows linearly with leverage, but your drag grows with the square of your volatility. So, as you increase the force that is multiplying your returns, the force that is eating them increases by that amount, squared. In other words, it was the exact same force that made the fund’s returns spectacular, that eventually led to an even more spectacular liquidation.

Again, not because the thesis was necessarily wrong, but because the financial structuring of the firm’s strategy maximised the wrong metric. It is a cruel mathematical mechanism that not even a crystal ball can save you from: Aschenbrenner’s fund was right for 6 months – 439% right to be exact – but the maths does not take kindly to leverage, so they only had to be wrong once.

Out with the new, in with the old

The 67% fall in July still left Situational Awareness up 80% for 2026, but it was enough to put thesis and overall direction to the side – bringing margin pressure to the forefront.2

And when margins called?

Citadel, in vintage Ken Griffin fashion, was on the other end of the phone, buying up positions at 90 cents on the dollar5 – many of which they have since offloaded and, as the FT reports, “made a killing doing” so.6

Footnotes

  1. Financial Times, A leaderboard of the biggest trading losses of all time (opens in a new tab), 17th August 2026. 2

  2. Forbes, The $35 Billion AI Trading Loss Exposes The Real Risk Of Leverage (opens in a new tab), 17th August 2026. 2

  3. CNBC, Why Situational Awareness hedge fund imploded, even in a tame stock market (opens in a new tab), 31st July 2026.

  4. EC Assets, Volatility Drag (opens in a new tab), undated.

  5. Financial Times, Citadel’s flagship fund surges 6% after Situational Awareness swoop (opens in a new tab), 5th August 2026.

  6. Financial Times, Citadel ‘makes a killing’ (opens in a new tab), 6th August 2026.