Segantii's Storage Chip Bet Blows Up, Citadel Takes Over Positions, Jane Street Loses $15 Billion
Nashnova编辑部
Three top hedge funds' Q2 13F filings reveal the positions behind July's blowup: SA went all-in on memory chips and imploded, Citadel's options matrix let it buy the debris at a discount, and Jane Street's direct investment in SA drove a ~$15 billion monthly loss — the firm's first negative trading month since 2016.
How concentrated was SA's memory-chip bet?
SA's Q2 portfolio hit $20.2 billion, up from $13.7 billion the prior quarter. Its top ten holdings made up 91.77% of total value — near-total concentration.
The two largest positions were SanDisk (SNDK) and Micron (MU), together accounting for over 56% of the book. This means → more than half the fund rode a single thesis: memory chips.
Micron was added at a rate of 27,712% quarter-over-quarter; TSMC at 11,713%. In plain terms = these were not incremental adds — they went from near-zero to max position in one quarter.
What did SA sell to fund the bet?
SA liquidated Nvidia, AMD, Intel, ASML, and Corning — every marquee AI-chip name it previously held.
It also sold VanEck's semiconductor ETF and put options on Nvidia, Broadcom, and AMD. This means → SA did not just rotate — it abandoned the "GPU-first" narrative entirely, pivoting to memory and compute infrastructure.
SA was founded in September 2024 by former OpenAI researcher Leopold Aschenbrenner. Its strategy: heavy long positions in AI infrastructure at up to 4× leverage, hedged by shorting legacy software stocks. First-half returns reached 439%, and AUM peaked at $45 billion.
What went wrong in July?
AI hardware and memory chips sold off sharply in July. Micron fell 28.69%; SanDisk dropped 46.57%.
SA's long book was crushed, but its software shorts — meant as a hedge — rallied instead. Both sides lost money at once. In plain terms = the "insurance" (short software) paid out to nobody.
Leverage amplified the drawdown. Margin buffers were breached, margin calls followed, and SA was ultimately forced to sell most of its public-market positions to Citadel at a discount to avoid forced liquidation.
How did Citadel end up on the other side?
Citadel Advisors held $88 billion in Q2. Its top five positions were all options — SPY calls, QQQ puts, QQQ calls, SPY puts, and Micron puts — forming a textbook long-short hedge matrix.
On both Micron and SanDisk, Citadel held calls *and* puts simultaneously. This means → it profited regardless of direction, while SA held only the long side.
That structure let Citadel buy SA's quality compute assets at fire-sale prices during the July crash. This reflects a basic divide: in extreme drawdowns, hedged portfolios survive and unhedged ones become inventory for those who are.
How did Jane Street get dragged in?
Jane Street's Q2 portfolio totaled $1.21 trillion, itself heavily hedged with SPY and QQQ options across both directions.
The fatal exposure was not proprietary trading — it was a direct investment in SA's fund. Per an internal Jane Street memo, SA's drawdown wiped out nearly all of the year's gains on that allocation.
Combined with losses on Asian non-AI equity longs and a broader reversal in previously profitable trades, Jane Street posted a ~$15 billion monthly loss in July — its first negative trading-revenue month since 2016.
What comes next?
Jane Street has closed most risk exposure in the areas that drove July's losses and scaled back risk-taking in other strategies. The loss will hit Q3 financials.
This means → Jane Street's Q3 report will be closely watched — the market wants to know whether the rest of the book can offset this hole.
The three 13F filings together teach one lesson: high leverage + high concentration + one-directional positioning = a lethal combination when the market turns. Citadel's hedged matrix survived; SA's all-in bet did not.
Content is for reference only, not financial advice.