With Warren Buffett’s retirement, Stanley Druckenmiller has cemented his status as one of Wall Street’s most closely watched asset managers. His Duquesne Family Office has unveiled its quarterly U.S. equity holdings as of March 31, revealing a striking strategic pivot in artificial intelligence investments that has drawn widespread market attention.
Druckenmiller fully liquidated his stake in Alphabet (GOOG, GOOGL), just one quarter after aggressively building up the position — a clear reversal of his earlier stance. The family office also sharply reduced its holdings in Amazon, as the two prominent AI application leaders were both trimmed from the portfolio.
Instead, Duquesne redirected capital toward the hardware ecosystem that powers AI operations. The investor argues the industry’s growth engine is shifting: enthusiasm is moving away from large-scale model training and toward commercial deployment via inference workloads, an area heavily reliant on memory, storage and custom semiconductors. This view underpins the firm’s latest portfolio reshuffle.
The new hardware positions fall into two distinct buckets. The first comprises memory and storage names. Surging demand from AI data centers has absorbed industry-wide production capacity, driving a cyclical uptick in component prices. Duquesne initiated stakes in Sandisk (SNDK), Micron Technology (MU) and Seagate Technology (STX), all of which have posted explosive financial results. Sandisk’s fiscal quarterly revenue more than tripled year-on-year, with its data center business jumping 233% quarter-over-quarter. Micron’s latest quarterly revenue nearly tripled, and the firm guides for over 200% revenue growth in the current period. Seagate, meanwhile, reported a 44% revenue rise last quarter and notched record profit margins.
Corporate executives have echoed signs of red-hot demand. Sandisk called the latest quarter a fundamental inflection point, noting multi-year agreements inked with high-value data center clients. Seagate added its highest-capacity hard drives are fully pre-sold through 2027.
The second focus is custom silicon. Druckenmiller opened new positions in semiconductor giant Broadcom (AVGO) and chip designer Arm Holdings (ARM). Broadcom’s AI accelerators serve as a popular alternative for major cloud providers, and the business is firing on all cylinders. Its AI revenue surged 106% year-on-year to $8.4 billion in the latest quarter, with management projecting robust growth to continue.
Retail investors are warned against mirroring these trades blindly, given considerable inherent risks. Regulatory requirements impose a six-week lag between quarter-end and public disclosure, meaning the March 31 holdings were only released in mid-May. During that interim period, share prices of these AI hardware plays rallied sharply, leaving late buyers facing far higher entry costs than Druckenmiller.
Memory and storage are famously cyclical sectors. While Sandisk, Micron and peers sport single-digit forward price-to-earnings ratios — which appear inexpensive on the surface — such valuations are typical at the peak of a business cycle. A shift in supply-demand dynamics could quickly erode product pricing and corporate profitability. Compounding the uncertainty, Druckenmiller has a history of locking in gains early; he once acknowledged misjudging the timing when he sold Nvidia prematurely. He has also long voiced skepticism over stretched valuations across the AI sector.
Alphabet, meanwhile, has delivered strong operating results following the stake exit. The company’s first-quarter revenue rose 22%, marking its 11th straight quarter of double-digit growth. Google Cloud revenue accelerated to a 63% year-on-year increase, and its cloud backlog nearly doubled to more than $460 billion. Operating income climbed 30%. Trading at a forward P/E of roughly 27, Alphabet’s valuation remains reasonable given its solid business trajectory.
In summary, Druckenmiller has rotated his AI portfolio from application-focused tech giants to underlying hardware, betting that inference infrastructure will lead the next phase of industry expansion. Even so, the disclosure lag, steep share price gains and cyclical headwinds facing hardware firms make this strategy ill-suited for blind copying. While AI hardware enjoys strong near-term momentum, cyclical swings remain a key threat. For their part, AI application leaders like Alphabet continue to deliver resilient growth. Investors are advised to conduct independent analysis before making investment decisions.
