All articles
    July 30, 2026 · 9 min read

    The Smartest Fund in AI Just Got Liquidated. Here's Why the Rebound Starts Now.

    Leopold Aschenbrenner's Situational Awareness sold its entire $13.7 billion public portfolio to Citadel in a single block trade before Thursday's open. The forced seller is gone. The shorts still need to cover. And the stocks he was holding just posted record earnings.

    At 9:30 AM Eastern on July 30, 2026, the most closely watched AI hedge fund in the world no longer owned a single public equity.

    Situational Awareness LP — the fund built by 24-year-old former OpenAI researcher Leopold Aschenbrenner — liquidated its entire public stock portfolio in one enormous block trade, reportedly to Ken Griffin's Citadel. Assets under management collapsed from a peak of $20-24 billion to approximately $10 billion overnight. A fund that returned 439% through June and was the best-performing large vehicle in global markets became a forced seller of everything it owned.

    The thesis was right. The leverage killed it.

    And the mechanics of what happens next could produce one of the sharpest rebounds in AI stocks this year.

    How a 439% Return Turned Into a Margin Call

    The story starts with conviction. Aschenbrenner left OpenAI in 2024, published "Situational Awareness" — arguably the most influential AI thesis of the decade — and launched a hedge fund that went all-in on AI infrastructure.

    The fund's public portfolio read like a StockGecko watchlist: SK Hynix, Nvidia, Micron, CoreWeave, Bloom Energy, Nebius, Lumentum, Coherent, plus Bitcoin mining plays like Riot Platforms, CleanSpark, HIVE Digital, and Bitdeer. The thesis was simple: AI needs chips, power, and data center infrastructure. Buy everything that supplies it.

    It worked spectacularly. Through June 30, the fund was up 439% net of fees, with year-to-date returns north of 147%. Leopold ran leverage as high as 4x — meaning every $1 of capital controlled $4 in positions.

    Then July happened.

    The AI semiconductor complex sold off 35%+ in three weeks. The KOSPI crashed 44% in 40 days. SK Hynix dropped 55% from its highs. CoreWeave, Bloom Energy, and Nebius were all hit simultaneously. And Leopold's short positions — bets against software companies like Adobe — moved sharply against him at the same time. The longs went down, the shorts went up, and 4x leverage turned a painful drawdown into an existential crisis.

    Goldman Sachs, JPMorgan, and Bank of America — the fund's prime brokers — began marketing the positions. The margin calls came. On July 24, Leopold wrote investors a letter calling the selloff "one of the best buying opportunities since early 2025" and invited fresh capital starting August 1. Six days later, the entire book was sold to Citadel.

    The Predatory Short Thesis

    Here's what makes this story more than just a cautionary tale about leverage.

    Leopold's positions were public. The fund's Q1 2026 13F filing showed approximately $13.68 billion in total notional exposure — every name, every size, visible to anyone who looked. When a fund running 4x leverage holds a concentrated, correlated book of AI infrastructure stocks, the playbook writes itself.

    Predatory funds — and the whisper on every trading desk is that several major ones were involved — saw the setup: a massive leveraged long in illiquid AI names, with known prime brokers, known margin thresholds, and known liquidation mechanics. The trade was straightforward: short the names Leopold was long, accelerate the drawdown, trigger the margin calls, and force the liquidation.

    Every tick lower in SK Hynix, CoreWeave, and Bloom Energy brought the fund closer to its margin limits. Every margin call forced more selling. Every forced sale pushed prices lower. The reflexive loop that made Leopold 439% on the way up worked in reverse on the way down — amplified by funds that were actively betting on his liquidation.

    This is not conspiracy. This is market structure. When a $20 billion fund's positions are public and its leverage is known, the liquidation becomes predictable. And predictable liquidations attract capital that profits from them.

    Citadel was reportedly the only firm with the balance sheet and risk appetite to buy the entire book in a single block. The reported price: roughly $0.40 on the dollar. Ken Griffin didn't buy Leopold's thesis at a 60% discount because he thinks AI is over. He bought it because he knows the forced selling is over — and the predatory shorts now have to cover.

    Why the Shorts Have to Cover

    This is the mechanical catalyst most coverage is missing.

    The funds that shorted Leopold's holdings into the liquidation have an open short position in every name he sold. As long as the forced seller existed, those shorts were profitable — the selling pressure from the margin calls pushed prices down, and the shorts rode the decline.

    But the forced seller is gone. Citadel owns the book now. Citadel is not a forced seller. Citadel is a patient holder with the longest time horizon and the deepest pockets on Wall Street.

    The moment the forced selling stopped, the dynamic flipped. The shorts still have open positions. The supply of selling they were riding behind has disappeared. If prices stabilize or bounce — which they already began doing after the Citadel deal was confirmed — the shorts face a choice: cover now at a modest profit, or hold and risk a squeeze as the AI names recover.

    History says they cover. SpotGamma noted that in prior cases where Citadel acquired forced-liquidation portfolios, the entry coincided with or preceded market stabilization in the affected sector. The acquisition removes the overhang of a forced seller, and the remaining short interest becomes the fuel for the rebound.

    If implied volatility and skew in Leopold's core names — Nebius, CoreWeave, SanDisk, Micron, the Bitcoin miner cluster — compress in the coming sessions, it confirms the flow was technical, not fundamental. And every short that was riding the liquidation becomes a future buyer as they cover into the recovery.

    The Fundamental Absurdity

    Step back from the mechanics and look at what the market just did.

    In the same week that Leopold was being margin called out of his positions:

    SK Hynix reported record profit. 62% global HBM market share. Forward P/E of roughly 5x.

    Samsung posted 19x earnings growth year-over-year. Record revenue. Stock ripping on the print.

    Microsoft beat consensus by 12%. Azure crossed $100 billion in annual revenue.

    Micron had already reported revenue up 345% year-over-year with EPS doubling consensus.

    The stocks Leopold was forced to sell are the most profitable semiconductor and AI infrastructure companies on earth — and they were sold at forced-liquidation prices into a drawdown driven entirely by leverage, not by any deterioration in the underlying businesses.

    Leopold himself acknowledged this. His July 24 letter called the selloff "one of the best buying opportunities since early 2025." He wasn't wrong about the thesis. He wasn't wrong about the opportunity. He was wrong about the leverage — and 4x exposure to a correlated book with public positions turned a buying opportunity into a forced exit.

    Forbes put it cleanly: "Forced liquidations driven by financing pressures rather than fundamental deterioration create unique opportunities for patient, unleveraged investors."

    The Setup for a Sharp Rebound

    Four conditions are now in place for a violent recovery in AI infrastructure stocks:

    First, the largest forced seller in the sector is gone. Leopold's $13.7 billion public book has been absorbed by Citadel in a single block. There is no more overhang from Situational Awareness.

    Second, predatory shorts need to cover. The funds that shorted into the liquidation are exposed to a squeeze now that the selling pressure they were riding behind has evaporated.

    Third, fundamentals are at all-time highs. Four consecutive record earnings from the largest companies in the semiconductor complex. $725 billion in committed hyperscaler capex. Supply sold out through 2027.

    Fourth, South Korea's government is actively intervening to stabilize the KOSPI. Emergency measures, stabilization fund activation, leveraged ETF curbs — all designed to put a floor under the market that was ground zero for the selloff.

    The combination of forced-seller removal, short covering, record earnings, and government intervention is the kind of catalyst stack that produces sharp, fast reversals. AI infrastructure stocks were already rebounding after the Citadel deal was confirmed. If the momentum continues, July 2026 will be remembered as the capitulation — not the beginning of a sustained decline.

    What This Means for On-Chain Traders

    Leopold's portfolio overlaps almost perfectly with the assets tracked on StockGecko. SK Hynix, Micron, Bloom Energy, CoreWeave, Nvidia, Riot Platforms — these all trade as equity perpetuals on Hyperliquid.

    The forced liquidation happened before Thursday's market open. On-chain traders on Hyperliquid could see the block trade being priced in real time through the perps while traditional investors were processing the CNBC headline. The AI infrastructure names began rebounding almost immediately after the deal was confirmed — and the Hyperliquid perps reflected that move 24/7.

    For anyone tracking AI infrastructure on StockGecko, the thesis is straightforward: the biggest forced seller in the sector just exited. The shorts are exposed. The earnings are records. The government is intervening. And the stocks are trading at single-digit forward P/E multiples after a 35-55% drawdown.

    The forced selling is over. The fundamentals haven't changed. The rebound is being fueled by the same mechanics that caused the crash — just in reverse.

    Track every name in Leopold's portfolio on StockGecko.


    Prices and data referenced as of July 30, 2026. This is not financial advice.