Former OpenAI Researcher’s AI Hedge Fund Implodes; Citadel Takes Over Portfolio

How a 20-Something AI Prodigy Built and Lost a Multi-Billion Hedge Fund

Leopold Aschenbrenner, a former OpenAI researcher still in his twenties, rocketed to Wall Street fame with his hedge fund Situational Awareness LP (SALP). Launched in 2024 with just over $200 million in seed capital from tech and quant insiders, the fund became a symbol of the AI boom, riding an ultra-concentrated bet on “AI infrastructure” to staggering returns: 439% in the year through June 2026 and a cumulative track record estimated in the thousands of percent, according to media reports.

That performance was built on a razor’s-edge portfolio. SALP put roughly 40% of its U.S. long equity book into cloud infrastructure company Nebius (NBIS), alongside heavy stakes in fellow AI hardware plays CoreWeave (CRWV) and SanDisk (SNDK). The fund also loaded up on Bitcoin miners and data-center power and cooling stocks, while using put options to short traditional semiconductor and software names. Leverage magnified the bets, with the fund taking on margin loans from multiple prime brokers.

When AI-linked stocks sold off sharply in July 2026, the concentrated longs cratered. At the same time, some of the legacy software names Aschenbrenner had shorted—such as Adobe—rallied, inflicting losses on the short book. The simultaneous hit on both sides drained collateral and triggered margin calls. Brokers demanded more capital or forced position reductions, ultimately pushing SALP into a mandatory unwind.

Rather than dump the massive equity block on the open market—which risked a disorderly crash—SALP negotiated a bulk sale of most of its listed stock and some hedges to Ken Griffin’s Citadel. The transaction, estimated in the billions of dollars, has not been commented on by either party. The episode marks a dramatic fall for one of the AI bubble’s most celebrated young guns.

The AI Supply Chain Bet That Backfired: Concentration, Leverage, and a Citadel Bailout

Nebius and CoreWeave: The Epicenter of the Bet

A 40% single-stock weighting in Nebius turned a high-conviction theme into a single-point-of-failure. When AI infrastructure names reversed in July, the overdependence on a few mid-cap proxies meant the fund had no cushion. The blow-up illustrates a structural risk that other hedge funds with similar “shovel and bucket” AI plays may now face: downstream liquidity evaporates fast when the narrative shifts, and prime brokers won’t tolerate breached margin limits indefinitely.

Betting Against Adobe and the Legacy Tech Basket

SALP’s put options on mature software and semiconductor companies were intended as a hedge and a stand-alone profit center. In the July turbulence, that hedge failed: shares of firms like Adobe actually rose, possibly as money rotated out of high-growth AI names into perceived quality. The double loss—big on longs, steep on shorts—compressed the fund’s equity base far more than a simple market-neutral book would have. It’s a textbook reminder that a poorly correlated hedge can amplify, rather than mitigate, drawdowns.

Leverage Multiplied the Pain

The presence of margin loans from multiple brokers was the fatal accelerant. As the mark-to-market value of its concentrated longs fell, the fund had to post additional collateral. When it couldn’t, the forced selling started, turning a controlled exit into a fire sale. The involvement of multiple prime brokers also likely complicated the unwind, with each acting in its own interest. The eventual block trade to Citadel—reportedly in the tens of billions of dollars—suggests the positions were so large that only a well-capitalized, risk-absorbing buyer could handle them without roiling markets further.

Citadel’s Opportunistic Move

For Citadel, the deal is a classic distressed-asset play. Acquiring the equity and hedges en masse likely came at a discount to quoted prices, giving it a potentially lucrative exposure to the same AI infrastructure names at a lower cost basis. Citadel can afford to hold and trade around the positions, whereas SALP’s leverage left it with no staying power. The transaction may also suppress near-term volatility in the affected stocks by taking supply off the public tape.

What the SALP Blow-Up Means for AI Portfolios and Hedge Fund Risk

• Watch for potential secondary selling in AI infrastructure names. Nebius, CoreWeave, and SanDisk were among SALP’s largest holdings. While Citadel has absorbed the initial block, any subsequent unwinding by Citadel—or by other funds that held similar positions—could create further downside pressure. Investors should track 13F filings and unusual block trades in these tickers.
• Reassess concentration risk in AI-themed portfolios. SALP’s 40% allocation to a single name proved unmanageable in a downturn. For allocators and fund managers, the episode underscores the danger of letting any single “story stock” dominate a portfolio, especially when the underlying liquidity is shallow and the thematic consensus is crowded.
• Beware of hedge books that break under rotation. The fund’s put strategy on legacy software failed to provide crisis alpha when the market rotated away from high-growth AI. Any hedged portfolio relying on a specific style or sector dispersion should be stress-tested for correlation breakdowns precisely at the moment they are needed most.
• The Citadel transaction signals a near-term floor but not an all-clear. By taking the inventory privately, Citadel may stabilize prices temporarily, but the original stress—a broader AI sector repricing—remains. Investors should monitor the trading volumes and price action of the affected names over the coming weeks for signs of lingering supply overhang.