Duquesne Family Office 13F Portfolio · Stanley Druckenmiller

The new dollars cluster into a handful of themes, and they are mostly second-derivative trades rather than the obvious headline names.

• Semis and the hardware underneath AI: Taiwan Semiconductor rose 19.1% to $281.6M and STMicroelectronics 18.8% to $232.4M, while AMD, Rambus and Qnity Electronics are all new to the book. Seagate's 140.6% increase to $117.7M is the tell — this is storage and memory, not GPU bravado.
• Megacap internet: Amazon's share count went up 1082.5% to $129.1M, and Alphabet is a brand-new $120.2M position. Both show only modest gains versus average cost, so Duquesne is paying up rather than harvesting.
• Owned media: Fox's two share classes arrived together, FOXA at $115.0M and FOX at $29.6M. Combined, roughly $144.6M is a bigger commitment than almost anything else in the portfolio.
• Transport and shelter: United Airlines more than tripled to $108.1M, Delta is new at $56.5M, and D.R. Horton, Cavco and Champion Homes all debuted. That is a rate-cut expression more than a stock-picking one.

The common thread is that none of these require the market's most crowded trade to keep working.

Almost every large trim sits on a substantial unrealized gain, which is what makes this look like funding rather than fear.

• Roku: shares cut 75.5%, releasing $78.2M, with a +91.7% gain versus average cost. The proceeds went straight into the two Fox classes — a swap out of streaming hardware economics into broadcast.
• Woodward: down 68.2%, $61.3M out, after a +153.7% run. This is an aerospace and industrial-controls business, not an energy company, despite its sector label.
• Teva Pharmaceutical: 73.7% lower, $59.4M released, at a +93.7% gain. That single cut explains a meaningful slice of health care's weight loss.
• Sandisk: trimmed 8.5%, but the remaining stake still carries a +2138.2% gain versus cost. When a winner is up that much, a trim is housekeeping, not a verdict.
• BBB Foods fell 6.7% and YPF 3.0% — small shaves that top up the larger bets.

Note what is missing: no mega-cap tech liquidation, no dumping of the semiconductor adds. Duquesne is selling winners it considers fully valued to buy winners it thinks are earlier.

The sector chart says consumer discretionary nearly doubled, from 7.36% to 14.56%, and that industrials went from 1.69% to 7.45%. Read carefully, because the NASDAQ labels are doing sloppy work here. Fox and Fox A are broadcasters — communication services businesses, not industrial ones — so most of the "industrials" jump is really a media bet. Roku's old "telecommunications" tag was never telecom either.

On the technology side, the weight climbed from 21.42% to 27.03%, and that understates things. Entegris is a semiconductor materials supplier filed under plastics, and Qnity is an electronics spin-off; both belong in the AI supply chain. The genuine compute-infrastructure exposure is meaningfully larger than the label suggests, and it is concentrated in memory, storage and analog rather than the GPU names everyone owns.

The real collapse is energy, 8.33% to 4.28%. Part of that is an artifact: Woodward's cut explains a lot of it, and Woodward was never an oil company. The other honest shift is finance, from nothing to 3.61%, which is entirely four crypto-linked miners. Net-net, a health-care book with a tech sleeve became a diversified compute-and-cyclical book with one huge medical anchor.

The through-line is a portfolio positioned for two things at once: a broadening AI capital-expenditure cycle that pays memory, storage, analog and IT distribution, and a rate-relief cycle that revives airlines and homebuilders. Both are second-derivative expressions, which is consistent with a manager who would rather own the picks and shovels than the headline.

The speculative sleeve is the interesting tell. Bitdeer at $64.7M sits 23.1% below average cost, Hut 8, Riot and Hyperliquid Strategies are all new, and together the digital-asset complex is roughly 3.6% of the book. That looks like an option on power and compute scarcity, sized so a wipeout would not matter.

The risk is concentration. Natera at 19.86% means one clinical or reimbursement headline can move the whole book, and health care at 33.65% still exceeds technology. If biotech risk appetite sours, both the anchor and the funding engine for these new positions weaken at the same time. The rotation is real, but it is being financed by a single name's gains — and that is the position to watch next quarter.