How Leopold Aschenbrenner, the ‘golden child’ of the AI trade, was laid low

By Ray Dombrowski · Reporting from Youngstown · July 31, 2026

You read headlines about "AI savants" and "superintelligence," but what matters to me is always the payroll.

When the Spreadsheet Runs Out of Numbers

You read headlines about "AI savants" and "superintelligence," but what matters to me is always the payroll. What matters is whether a bet on a chip manufacturer or a data center actually puts names on a paycheck that lasts longer than six months. This week, we watched Leopold Aschenbrenner’s Situational Awareness hedge fund get forced to unwind its entire public stock portfolio. It was less of a market correction and more of a structural failure—a spectacular bonfire built entirely out of hype.

The details are textbook for a speculative bubble bursting. Aschenbrenner, the former OpenAI researcher who graduated valedictorian from Columbia, launched Situational Awareness after his departure from OpenAI in 2024. He marketed the fund on the premise that AI required an "enormous expansion of computing power, advanced semiconductors, memory and energy infrastructure." The numbers were dizzying: the fund reportedly grew to $45 billion at one point. But the underlying bets—heavy stakes in companies like Nebius Group, Sandisk, Micron, and CoreWeave—were fundamentally unsound.

Prime Brokers Step In When Margin Calls Hit

The mechanics of the collapse are laid bare by the reporting. According to CNBC, Situational Awareness’s prime brokers—Bank of America, Goldman Sachs, and JPMorgan Chase—had to rush to raise cash just to meet margin requirements. The fund sustained "significant losses" as its AI infrastructure bets stumbled while short positions in software companies moved sharply against it.

The sheer scale of the loss was staggering. By the time ibtimes.com reported on the forced unwinding, multiple core holdings had lost more than 35% of their market value in a single month. The fund’s initial trajectory—running up allegedly 439% net of fees from January 1, 2026 through June 30, per protos.com—was built on the unsustainable inflation of valuations. It was all predicated on future demand for electricity and chips, not current employment figures or stable revenue streams.

The Ghosts of the Dot-Com Bubble Return

This isn't new. Anyone who has spent time looking at capital flow knows this pattern. What we are witnessing is a perfect echo of the dot-com bubble burst. In both cases, massive amounts of venture capital and investment money pour into a newly hyped technological sector—the internet in the late nineties; AI infrastructure today. Valuations inflate to absurd heights because the promise (the "what if") vastly outstrips the reality (the actual cash flow).

The shared mechanism is clear: unsustainable inflation leading to forced deleveraging and systemic capital flight. The Nasdaq’s rise by 600% between 1995 and its March 2000 peak, followed by a 78% fall by October 2002, provides the template. Aschenbrenner's fund didn't build on stable economic growth; it built on narrative momentum. When that momentum stalls, the capital flees faster than any labor market can absorb it.

The lesson isn’t about AI being bad; it’s about finance treating technological potential as guaranteed profit. The money doesn't follow the utility; it follows the hype cycle. And when the hype runs out of steam, the spreadsheet always shows a deficit that no amount of conviction or academic brilliance can fix.

How Leopold Aschenbrenner, the ‘golden child’ of the AI trade, was laid low

When the Spreadsheet Runs Out of Numbers

Prime Brokers Step In When Margin Calls Hit

The Ghosts of the Dot-Com Bubble Return

Sources