The legendary investor Stanley Druckenmiller just sent a clear signal through his latest 13F filing: the surest way to profit from AI isn't selling the picks and shovels, but owning the miners. ๐
His firm, Duquesne Family Office, completely exited positions in five major chip and photonics companies, including Micron, Intel, and Broadcom. In the same quarter, it aggressively built massive new stakes in the two biggest corporate spenders on AI infrastructure: Amazon (AMZN) and Alphabet (GOOGL).
The filing shows a dramatic portfolio shift. Reported holdings jumped from $3.4 billion to $5.2 billion, and the position count grew from 70 to 95. While a 13F is just a snapshot in time, the sheer scale of this rotation demands a closer look.
The New Core of the AI Trade: Buyers Over Builders
The contrast in Duquesne's new positions is stark. Let's break down the two major additions that define this new strategy.
1. Amazon (AMZN): A Position Rebuilt Over 1,000% ๐
Duquesne didn't just add to Amazon; they rebuilt the position from the ground up. The firm increased its stake from 45,800 shares to 541,600 sharesโan increase of more than 1,000%. This stake is now worth roughly $129 million.
Even more telling is the addition of call options on another 459,300 shares, representing over $100 million in underlying value. Combined, this puts Amazon among the portfolio's largest bets.
Why it matters: Amazon's Q2 earnings backed up this conviction. AWS revenue growth accelerated to 37% YoY, the fastest rate in 18 quarters. More importantly, CEO Andy Jassy raised 2025's capital spending plan to a staggering $220 billion, up from $200 billion. This is the fuel for the AI fire, and Amazon controls the pump.
This dramatic shift has split the investment community. Is Druckenmiller early in seeing a bubble, or is he simply ahead of the curve? Let's look at both sides of the debate.
2. Alphabet (GOOGL): A Brand-New $120 Million Bet ๐จ
Alphabet wasn't in the portfolio at the end of Q1. By June 30, Duquesne had accumulated 336,300 Class A shares, worth about $120 million, representing 2.3% of the portfolio.
This new position aligns perfectly with the Amazon thesis. Alphabet's Q2 capital expenditures doubled YoY to $44.9 billion, and the company raised its full-year plan to a range of $195 billion to $205 billion.
The Core Thesis: These two companies aren't just buying chips; they are designing their own. Amazon's in-house chip business is already at a $25 billion annual run rate, and Alphabet's cloud unit sells systems built on its own TPU chips. They hold the demand and the design.
๐ In-Depth Fundamental Analysis
| Company | Share Price | P/E Ratio | P/B Ratio | ROE | Operating Margin (OPM) | Revenue Growth |
|---|---|---|---|---|---|---|
| AMD (Advanced) | $466 | 118.98 | 11.32 | 10.20% | 17.25% | 50.10% |
| AMZN (Amazon.com,) | $266 | 21.39 | 5.20 | 30.56% | 13.69% | 19.60% |
| AVGO (Broadcom) | $362 | 60.41 | 19.67 | 37.28% | 48.99% | 47.90% |
| COHR (Coherent) | $287 | 69.94 | 5.16 | 7.98% | 11.83% | 33.70% |
| GOOG (Alphabet) | $342 | 17.15 | 6.71 | 48.68% | 34.03% | 24.20% |
| GOOGL (Alphabet) | $345 | 17.30 | 6.77 | 48.68% | 34.03% | 24.20% |
| INTC (Intel) | $93 | 0.00 | 5.35 | -10.71% | 12.19% | 25.40% |
| LRCX (Lam) | $307 | 53.24 | 30.82 | 65.07% | 37.39% | 30.00% |
| LSCC (Lattice) | $118 | 470.72 | 20.50 | 4.93% | 13.33% | 62.20% |
| MU (Micron) | $937 | 21.19 | 10.50 | 66.64% | 80.37% | 345.70% |
| STM (STMicroelectronics) | $50 | 98.84 | 2.55 | 2.69% | 6.88% | 26.10% |
| TSM (Taiwan) | $412 | 30.87 | 84.74 | 39.97% | 60.34% | 36.00% |
Scenario Analysis & Conclusion
This move suggests Druckenmiller believes the 'picks and shovels' phase of the AI trade is maturing, and the durable value is shifting to the companies with the end-customer demand and the balance sheets to dictate terms.
| Scenario | Likelihood | Impact on AMZN & GOOGL | Impact on Sold Chip Stocks (e.g., MU, INTC) |
|---|---|---|---|
| Best Case | Medium | AI spending accelerates, driving cloud growth and ad efficiency. Stocks see multiple expansion. | Commoditization increases; they lose pricing power to in-house designs. |
| Worst Case | Low | AI spending pauses, leading to overcapacity in cloud and a slowdown in ad revenue. | A pause in spending could lead to a sharp correction, but they are already sold. |
Conclusion: While a 13F is a lagging indicator, the message is coherent. Druckenmiller isn't bearish on AI; he's repositioning for the next phase. He sold the suppliers to buy the kingpins. For investors, this suggests that focusing on the companies with the scale to fund and consume the AI build-outโlike Amazon and Alphabetโmight be the more durable long-term play. As always, this is not financial advice, and positions can change quickly, so conduct your own research before making any investment decisions.
