Before Nasdaq's Darkest Day, Institutional Giants Had Quietly Exited Chips in Q3
nashnova research
Thirty-six days before the Philadelphia Semiconductor Index plunged 12.25% in a single session, 13F filings showed three top hedge funds splitting sharply — Druckenmiller nearly exited chips entirely, Lone Pine reshuffled within the sector, and Tepper doubled down on Nvidia and Micron — three opposing bets in the same window, exposing a fundamental disagreement on the AI-chip cycle.
What did Druckenmiller cut — and what did he keep?
TSMC holdings dropped from 589,680 shares to 32,700 — a 94% reduction. Quarter-end market value shrank from roughly $282 million to about $15 million.
ARM, AMD, Lam Research, SanDisk, STMicroelectronics, Entegris, and Rambus were all liquidated. Chip-and-storage exposure fell from 16% to 10% of the portfolio.
The remaining chip-adjacent bet concentrated in Seagate — a hard-drive maker — with holdings raised to about $315 million plus $90 million in call options. He simultaneously opened three gold-mining positions: Newmont, Barrick Gold, and Agnico Eagle, totaling roughly $129 million.
This means → Hard drives plus gold is perhaps the most "de-AI'd" portfolio configuration available. Druckenmiller's read: the chip cycle has peaked; rotate into physical assets.
Lone Pine reshuffled rather than exited — so why did it lose even more?
Flagship fund Cypress posted a 44% year-to-date gain by end of June, then gave nearly all of it back — finishing Q3 at just 0.6% YTD, a single-quarter drawdown of roughly 30%. October added another 8.8% loss.
Sell side: Applied Materials — a semiconductor-equipment maker — was cut by 92%, from $873 million to $48 million. ASML was trimmed 25%, Teradyne 34%. Seagate, Corning, and TTM Technologies were fully liquidated.
Buy side: TSMC holdings surged from 101,242 shares to 1.72 million (about $783 million). Intel was a new position at roughly $721 million. Excluding Seagate, core chip exposure at quarter-end was about $2.72 billion — virtually unchanged from $2.78 billion the prior quarter.
In plain terms = Lone Pine didn't shrink its total chip bet — it swapped which horses it backed. But the reshuffle itself was poorly timed, likely one reason it kept bleeding 8.8% in October.
What was Tepper betting on in the other direction?
Micron holdings rose 28% to 1.25 million shares, making it the fund's single largest position at roughly $1.33 billion. Nvidia holdings increased 31% to 2 million shares.
Lam Research and ASML were added to; Applied Materials was a new position — the very stock Lone Pine was dumping.
This means → Tepper's call was the exact opposite of Druckenmiller's: the chip cycle isn't over, and the pullback is a buying window.
What does the bigger picture look like?
Goldman Sachs data: tech long-short funds peaked at a 24% YTD gain in June. High-beta tech momentum stocks — the most market-sensitive names — had rallied 125% from the end of Q2.
After the semiconductor index's "Black Thursday," those same momentum names fell 56% from their highs, nearly erasing the full year's gains.
Goldman's John Flood told Reuters that these funds had been "selling the very mega-cap tech stocks they'd bought just the day before."
This reflects a market where institutional consensus is nowhere close to forming — Druckenmiller chose to leave early, Lone Pine chose to reshuffle internally, Tepper chose to press the bet. Which call proves right won't be clear until Q4 filings are disclosed.
市场有风险,内容仅供研究参考,不构成投资建议。
