AI Stock God Fund Q3 Plummets 92%: After the Margin Call, He Uses Options to Bet Again

Nov 17, 2026 17:09:00

Author: Hu Tao, ChainCatcher

On November 16, former OpenAI researcher Leopold Aschenbrenner's hedge fund Situational Awareness LP submitted a 13F quarterly holdings report to the U.S. SEC for the period ending September 30, 2026. The numbers are staggering: the nominal value of publicly held positions plummeted from $20.24 billion at the end of the second quarter to about $1.58 billion, a quarterly shrinkage of 92.2%.

This document has caused market turmoil not only because of the drop itself but also because it reveals the story after the liquidation in July—Aschenbrenner did not pull back but instead doubled down on the options betting table.

From 439% to -67%: Completing a Cycle in Four Months

To understand the impact of this 13F, we need to turn the clock back six months.

Leopold Aschenbrenner, 25 years old this year, gained fame in 2024 for publishing a 165-page long article titled "Situational Awareness," in which he predicted that AGI would arrive between 2027 and 2030. After being fired from OpenAI the same year, he founded Situational Awareness LP with $225 million in seed funding provided by Silicon Valley luminaries such as Stripe founders the Collison brothers and former GitHub CEO Nat Friedman.
The fund's strategy is clear and extreme: All in on AI infrastructure—storage chips, data centers, clean energy—while simultaneously shorting software stocks like Adobe. In the first half of 2026, the fund's return reached as high as 439%, and its management scale ballooned to about $45 billion, making it Wall Street's hottest "AI stock god."

However, this brilliance was built on a leverage ratio of about 400%. In July, the AI infrastructure sector collectively corrected, and software shorts faced a reverse squeeze. The dual pressure combined with high leverage triggered margin calls. Aschenbrenner was forced to sell most of his publicly held stock portfolio at a discount to Citadel, resulting in a net worth drop of about 67% that month.

In a letter to investors, he admitted that the fund was "closer to permanent capital loss than acceptable," but insisted that this was a "costly yet invaluable lesson" and announced that all leverage had been cleared.

The Truth Revealed by Q3 13F

If the August Q2 13F was "the last holdings chart before the liquidation," then the November Q3 13F is "the first complete snapshot after the liquidation"—it proves that Aschenbrenner did indeed clear the leverage, but he did not reduce his risk appetite; he simply switched to a more covert leverage tool.

The specific changes are as follows:

Liquidated 21 positions, including once absolute heavyweights: Sandisk (previously $5.67 billion), Micron ($5.57 billion), Bloom Energy ($1.9 billion), TSMC ($1.27 billion), Nebius ($1.23 billion), and all positions in the computing power supply chain such as CoreWeave, Core Scientific, and STMicroelectronics were reduced to zero.

Reduced holdings in 3 positions: mining company IREN cut by 94%, CleanSpark reduced by 75%, Bitdeer reduced by 44%.

New builds and increases: Bloom Energy call options surged about 10 times to a nominal value of $389 million; SharonAI (SHAZ) increased by 50% to about $441 million, becoming the current largest single holding (27.9%); simultaneously, new call option positions were established for Nebius, Sandisk, and Roundhill Memory ETF, totaling about $656 million.

Structural upheaval: Currently, the nominal value of call options in the portfolio totals $1.045 billion, accounting for 66.23%; the underlying stocks are only $533 million, accounting for 33.77%. This is a long position ledger primarily based on options—something that has never appeared in the previous four quarterly reports.

What does this mean? Aschenbrenner has not abandoned his faith in AI infrastructure, but he has given up on expressing that faith through underlying stocks, opting instead for deep out-of-the-money call options to seek high-leverage returns. From total return swaps (TRS) to call options, the tools have changed, but the essence remains the same: he is still betting in one direction with other people's money (although the option premiums are less than the nominal value, gamma risk can also be deadly when volatility amplifies).
Silicon Valley's "hero narrative" versus Wall Street's cold joke

After the July liquidation, Silicon Valley's reaction was starkly different from Wall Street's.

Sequoia Capital partner Pat Grady stated in a Bloomberg TV interview: "Our judgment is that he will be an important figure in Silicon Valley for a long time." Veteran investor Elad Gil publicly announced his first application to invest in his fund; Redpoint Ventures managing director Logan Bartlett bluntly said: "There is a hero archetype here. Leopold was punched, and it instead sparked unity among everyone."

Wall Street's attitude, however, was much colder. S3 Partners founder Bob Sloan pointedly stated: "Let's be clear, this is a super concentrated position, a super crowded position, and also a super high-leverage position." Wall Street trading coach Jerry Diao's assessment was more straightforward: "Many people see this as 'it's not a question of whether it will blow up, but when it will blow up.'"

New York University Stern School of Business professor Gygmy Gonnot provided a structural explanation: "Silicon Valley rewards those who make correct judgments in transformative technology directions, while Wall Street rewards those who create considerable risk-adjusted returns while preserving principal."

The conflict between these two evaluation systems was brutally validated in the Q3 13F: Aschenbrenner's directional judgment on AI infrastructure may not have been wrong—Sandisk and Micron's fundamentals indeed did not collapse after July—but he did not survive to see the day when logic materialized. The Q3 options ledger indicates that he is still trying to recover what has already been lost using the same logic and higher-risk tools.

A Problem That Needs to Be Addressed

From an investment perspective, the most dangerous signal in the Q3 13F is not the "92% shrinkage," but the "66% options."

The 13F document's disclosure of options only shows nominal value (strike × contracts × 100), without showing strike prices, expiration dates, or actual premium costs. This means that outsiders cannot determine whether these call options are deep out-of-the-money or at-the-money, nor can they know when the premiums will go to zero if the underlying continues to fall.

However, tracking data from Tracefour shows that the underlying assets corresponding to these new options positions generally continued to decline in Q3: SharonAI fell from Aschenbrenner's entry price of about $68.59 to $22.56, a drop of 67%; Bitdeer fell from $11.57 to $5.60, a drop of 51%. If the strike prices of its call options are close to the entry costs, then the Q3 options positions may have suffered losses equal to or even more severe than the underlying stocks—because options have time value decay.

SharonAI blew up in July with 400% leverage, then rebuilt positions with call options in August-September, while Q3's AI infrastructure stocks continued to decline, and these options may have significantly depreciated in value by the snapshot on September 30. The $1.58 billion book value may already be a number "after the drop."

Conclusion

Aschenbrenner's story is worth the attention of the crypto industry not only because he is a phenomenal figure but also because his predicament is highly similar to many participants in this industry: an unwavering faith in technological direction, coupled with a serious lack of respect for risk management.

From $225 million in seed funding to a peak of $45 billion, and then to $1.58 billion in the Q3 13F, Situational Awareness LP has completed a full leverage cycle in 16 months. The Q3 options ledger shows that this cycle may not be over—it has simply switched from "total return swaps" to "call options," from one form of leverage to another.

Silicon Valley can continue to applaud the "hero archetype," but the numbers on the 13F do not lie. When 66% of a holdings report consists of call options and the largest holding is a small-cap stock that has dropped 67%, "the direction is correct" is no longer a valid defense.

The Q4 13F document, expected to be disclosed in February next year, will tell us whether these call options are exercised, closed, or have gone to zero. At that time, whether the "AI stock god" has been reborn from the ashes or has completed a second liquidation will truly be revealed.

Recent Fundraising

More
Nov 17
$9MNov 16

New Tokens

More
Dec 31
Dec 31
SSaturnSTRN
Dec 31

Latest Updates on 𝕏

More
SSolanaFollowedAbra
Nov 16
BBaseFollowedMultipli
Nov 16