76%
Aswath Damodaran (Musings on Markets) · 2026-08-11 · original: EN

It was not the AI call that broke the fund

What the market gives easily, it also takes away just as easily, and if you put into place strategies that are designed to deliver outsized returns, you have to live with the reality that you can have outsized losses.

I am not as convinced that investing conviction is a net plus.

Aswath Damodaran (Musings on Markets) · 2026.08.11

The fund did not break because the call that AI would win was wrong. It broke because of how that call was carried, in a few positions and with borrowed money. That is the argument Damodaran makes about the July collapse.

That report puts the scale at $45 billion of assets at the start of July, down to about $10 billion a month later. The firms holding the collateral underneath were Bank of America, Goldman Sachs and JPMorgan Chase.

The two channels conviction runs through

The conviction Damodaran describes is a procedure rather than a personality. It is the state of believing strongly in three things at once: that an asset is mispriced, that the market will correct it, and that the correction arrives inside the period you plan to hold it. Believe all three and conviction turns into money in only two places. One is how much goes into a single position, the other is how much debt sits on top of that money. The same judgment produces a different outcome once those two are set large. He also points at the fee structure. Charging 2% of assets and 20% of gains is a handicap a long term investor struggles to clear, and on the manager's side it is a reason to carry more risk.

The top five holdings were over 76%

Bloom Energy 22.8%SanDisk 18.8%CoreWeave 14.4%The other 21 positions 44.0%

Across 26 disclosed US long positions, the top five together came to more than 76%. This is what it looks like when conviction cuts down the number of names.

SpotGamma tally of 2026-07-30, based on disclosed holdings · retrieved 2026-08-11

That tally reports gross exposure reaching four times the fund's own capital. Cut the number of names on one side and add debt on the other, and a small move in price puts all of the forced selling in one place.

What July 2026 tested was not the judgment

Damodaran does not accept the explanation that runs through age.

The common reading

A 25 year old had no experience. He broke because he lacked the wisdom that supposedly comes with years.

VS
Damodaran's reading

Age explains nothing. The path to the break was conviction translating into fewer names and more debt.

One side of the same collapse looks at the person, the other looks at the channel through which conviction moves into capital. So what to check in your own portfolio is not the strength of the judgment either. It is two things: what share one position takes of the whole, and whether borrowed money is sitting on that position. With both small, a wrong call still leaves room for the next one. With both large, a right call can still be sold out from under you. What went on trial in July was not the outlook for AI but the size of the vessel that outlook was poured into.

In three lines
  • Damodaran has written about the July collapse of a hedge fund that concentrated on artificial intelligence infrastructure.
  • On the surface it reads as an argument over whether the call that AI would win was wrong.
  • What he actually examines is not whether the call was right, but the path by which conviction gets translated into fewer positions and more borrowed money.

Sources

  1. Original Aswath Damodaran, Musings on Markets · 2026-08-11
  2. Report Quartz · 2026-07-31 · source for the drop in assets and the prime brokers
  3. Tally SpotGamma · 2026-07-30 · source for position weights and gross exposure

Retrieved 2026-08-11 · position weights follow SpotGamma's 2026-07-30 tally of disclosed holdings, and the asset figures follow Quartz reporting of 2026-07-31.

This author's record

3posts0directional calls
See the full record →