The Fund That Was Surest About AGI Is the One That Broke

I went looking for the receipts on this one. Reddit threads from the last 30 days, credible discussion, anything with a number attached — nothing held up. So I can’t tell you which fund lost what, and I’m not going to pretend otherwise.

What I can do is explain why the AGI trade keeps ending this way. Not because the thesis is wrong. Because of how the thesis gets converted into a position.

“AI Exposure” Is Actually Three Different Bets

Everyone says they’re long AI. That phrase hides at least three bets with completely different risk profiles.

The first is compute demand — GPUs, data centers, power infrastructure. The second is model capability: the assumption that the next generation is meaningfully smarter than the last. The third is monetization, where that capability turns into enterprise revenue.

These do not move together. Compute demand already showed up in the numbers. Capability has kept climbing. Monetization is the laggard, and it isn’t close. That gap is where the pain lives, and it’s why “I was right about AI” and “I made money on AI” can both be true statements about different people.

Conviction Concentrates Portfolios

Leopold Aschenbrenner’s Situational Awareness, published in June 2024, became one of the most-cited documents in the industry almost immediately. AGI around 2027, then a fast takeoff. Clean narrative, specific dates, and — this is the part that matters — an implicit shopping list. If you believe it, what to buy follows automatically.

A clean narrative is a liability in a portfolio. High conviction pushes leverage up and hedges down. Once you decide something is certain, paying for the other side of the trade feels like a waste of capital. That’s the moment the entire book gets staked on a single scenario.

And everyone who read the same document buys the same names. The industry term is crowded trade. The door looks wide on the way in. It gets narrow on the way out, and it gets narrow for everyone simultaneously.

Being Right About Direction Won’t Save You

Suppose AGI does arrive. You can still get wiped out.

Any leveraged position has a clock attached, and the clock is called a margin call. A 20-30% drawdown liquidates you regardless of what happens next. The recovery belongs to whoever was still solvent to see it. Keynes said it best: the market can stay irrational longer than you can stay solvent.

The dot-com bust is the clean version of this. “The internet changes everything” was completely correct. People who levered into that correct thesis in 1999 were gone by 2002. Amazon fell roughly 90% and then became Amazon — but only for the shareholders who never got margin-called out of it. Being right and getting paid are separate events.

The Real Risk Is Repricing, Not a Crash

AI valuations already have future cash flows baked in. The question isn’t whether those flows exist — it’s when they land. Timing assumptions function as the discount rate here.

Price AGI at 2027 and today’s multiples make sense. Push it to 2032 and the thesis survives intact while present value gets shredded. The market doesn’t need a scandal. It doesn’t need anyone to prove AI was a fraud. It just needs a quiet consensus shift toward “this is running about three years late.”

That shift never announces itself. It starts with one line in an earnings call about a customer deferring an AI budget to next quarter. Then another. Boring inputs, violent outputs.

This Scales Down to Your Brokerage Account

It reads like a hedge fund problem. It isn’t.

“Is this narrative persuasive?” and “how much should I bet on it?” are separate questions, and the danger starts the instant you translate conviction directly into position size. However sure you are, the position has to be small enough that being early doesn’t end you.

It’s also worth auditing where your view came from. Did you buy this name because of work you did, or because you read one very well-written document? If it’s the second, understand that tens of thousands of people read it too — and you’ll be reaching for the sell button at the same moment they do.

Direction or Timing — Which One Are You Long?

I think AI reshapes the economy. I also think the thesis being right and the trade being profitable are unrelated propositions, because markets price timing, not just direction.

So check what you actually own. A bet on “AI is coming” is a different instrument from a bet on “AI arrives by 2027.” If it’s the second one, run the math on a three-year delay and see whether your account survives the wait.