$15 billion. In one month.
Yesterday, news broke that Jane Street — one of the largest and most sophisticated trading firms in the world — took a roughly $15 billion hit in July.
Think about that number for a second.
$15 billion.
And this is Jane Street. A firm known for quantitative trading, technology and risk management. Despite this extraordinary loss, Jane Street has still generated more than $40 billion in trading revenue this year.
So how does a firm like Jane Street lose $15 billion in a month?
Part of the answer leads to one of the most fascinating stories I have seen during this AI boom.
Situational Awareness.
First, meet Leopold Aschenbrenner
Leopold Aschenbrenner is a former OpenAI researcher who became well known in the AI community after publishing Situational Awareness: The Decade Ahead.
His thesis was extraordinarily bullish on AI.
AI capabilities would advance rapidly. The amount of compute required would explode. And enormous amounts of physical infrastructure — GPUs, memory, networking, data centers and power — would have to be built to support it.
He didn't just write about the thesis.
He created a hedge fund around it.
And called it Situational Awareness.
The fund made highly concentrated bets on companies positioned to benefit from the AI infrastructure boom.
And those bets worked.
Wall Street noticed. One of the sophisticated investors that invested in the fund was Jane Street.
Then July happened
AI-related stocks began selling off.
Situational Awareness's highly concentrated and leveraged positions suddenly started working in the opposite direction.
By the end of July, the portfolio was in a free fall. An extraordinary reversal.
But here's the part of the story I find particularly interesting.
The AI thesis didn't suddenly disappear
AI didn't disappear in July.
The world didn't suddenly decide it didn't need GPUs.
The long-term demand for AI compute didn't suddenly go to zero.
Situational Awareness may ultimately prove to have been right about the direction of AI.
But being right about a technology transformation and being right about the investments surrounding that transformation are not necessarily the same thing.
And then we get back to Jane Street
Jane Street was an investor in Situational Awareness and also had exposure to other AI-related technology positions.
When July's selloff hit, the result was approximately $15 billion in losses in a single month — an extraordinary number even for a firm of Jane Street's scale.
But I think the story tells us something else.
The financial bets surrounding AI are becoming enormous
I believe AI is one of the most important technology transformations I have seen during my career.
But I keep coming back to the economics surrounding it.
In my first article, I questioned whether the returns will justify the hundreds of billions being invested in AI infrastructure. In my second, I looked at AI compute potentially becoming an investable asset.
And now we have Situational Awareness. Another piece of the puzzle. Enormous amounts of capital are being placed behind one basic assumption:
AI usage will explode, driving enormous demand for compute and the infrastructure behind it.
Maybe that proves absolutely correct.
But eventually, somebody has to pay the bill.
And ultimately, that's the customer using AI.
Enterprises still have to turn all this extraordinary technology into measurable economic value.
That's the part of the equation I keep coming back to.
Which leaves me with a bigger question
The Situational Awareness story shows just how much financial expectation is now being built around that thesis.
How much of that expectation ultimately depends on enterprises generating the enormous economic returns everyone assumes are coming — and generating them quickly?
Perhaps the biggest risk to the AI investment boom isn't that the technology doesn't work.
Perhaps the technology works extraordinarily well.
But what if the economics take longer to catch up with the capital markets' expectations?
That's the question I want to explore next.
I hope you're enjoying reading Inside Enterprise AI as much as I'm enjoying writing it.