Stanley Druckenmiller’s Duquesne Family Office LLC sold every share of Alphabet Inc. Class A it held and slashed its Amazon.com Inc. position by roughly 94 percent during the first quarter of 2026, according to SEC filings. The firm simultaneously built new positions in semiconductor companies Broadcom and Arm Holdings. The trades represent one of the sharpest rotations away from large-cap internet platforms and toward AI chip suppliers by a high-profile investor this year.
Why Druckenmiller’s exit from Alphabet and Amazon signals a strategic shift
Duquesne’s quarterly filings tell a clear before-and-after story. As of December 31, 2025, the firm reported holding 385,000 shares of Alphabet Inc. Class A and 737,940 shares of Amazon.com Inc., figures detailed in the family office’s prior 13F report. By March 31, 2026, the Alphabet position had disappeared entirely from the holdings table, and Amazon had been cut to just 45,800 shares with a reported value of $9,539,000.
The timing carries weight. Druckenmiller built his reputation on concentrated bets that reflect strong directional conviction. Exiting a position completely, rather than trimming around the edges, suggests the firm sees better risk-adjusted returns elsewhere. In this case, the destination is clear: companies that design and manufacture the chips powering artificial intelligence infrastructure.
The hypothesis behind the trade is straightforward. If hyperscale cloud providers and enterprise customers continue spending heavily on AI training and inference hardware, semiconductor makers capture a growing share of those dollars at high margins. Alphabet and Amazon, by contrast, face a different calculus. Both companies are themselves massive buyers of AI chips, meaning their capital expenditure rises alongside the very trend Druckenmiller appears to be betting on. Selling the buyers and owning the suppliers is a way to position on the same side of AI spending growth without absorbing the cost burden.
Duquesne’s filings also show that the rotation did not occur in isolation. Other large technology and consumer holdings were adjusted during the period, but the complete sale of Alphabet and the near-elimination of Amazon stand out in both size and direction. The move effectively removes two of the most widely held U.S. megacap stocks from a portfolio that had previously leaned heavily on internet advertising and e‑commerce growth.
SEC filings show Broadcom and Arm replacing big tech in Duquesne’s portfolio
The Q1 2026 13F filing, submitted on May 15, 2026, lists new positions that did not appear in the prior quarter’s report. Duquesne held 195,955 shares of Broadcom and 106,700 American depositary shares of Arm Holdings as of March 31, according to the detailed position table. Both companies sit at the center of AI chip design: Broadcom supplies custom accelerators and networking silicon to major cloud operators, while Arm licenses the processor architecture used in an expanding range of AI-capable chips.
The shift in portfolio composition is stark. Where Duquesne’s year-end 2025 filing showed heavy exposure to consumer internet revenue through Alphabet and Amazon, the first-quarter 2026 filing tilts toward hardware suppliers whose revenue growth depends on sustained capital spending by those same internet giants and their enterprise customers. The 45,800 Amazon shares that remain represent a fraction of the original stake, roughly a 94 percent reduction from the 737,940 shares reported three months earlier.
On a dollar basis, the new semiconductor positions more than offset the value of the residual Amazon holding. That imbalance underscores the degree of conviction behind the pivot. Rather than simply diversifying into chips at the margin, Duquesne appears to be reorienting a significant portion of its technology exposure toward companies that sit deeper in the AI infrastructure stack.
No accompanying commentary or letter came with either filing. Form 13F disclosures are limited to share counts and market values at quarter-end, and they do not explain the investment thesis or subsequent trades. Still, the pattern across Duquesne’s recent regulatory submissions points to a consistent theme: reducing dependence on consumer-facing platforms and increasing exposure to the components and architectures that enable AI workloads.
What the rotation says about AI, capital spending, and risk
For other investors, the most important takeaway may be what this rotation implies about the next phase of AI-driven spending. If Druckenmiller is right, the more durable opportunity lies with firms that sell critical inputs into data centers, rather than with the platforms that sit on top of that infrastructure. That view favors chip designers, networking specialists, and other hardware suppliers that can translate rising AI demand into higher pricing and volumes.
At the same time, the trades highlight a view on risk. Large-cap internet stocks have already rerated higher on AI optimism, leaving less room for error if growth slows or regulatory pressure intensifies. Semiconductor names, while volatile, offer more direct leverage to each incremental dollar of AI capital expenditure. By rotating from Alphabet and Amazon into Broadcom and Arm, Duquesne is effectively swapping platform risk for infrastructure risk – a bet that the build-out of AI computing capacity will remain robust even if end-market winners shift over time.
Whether that judgment proves correct will take several quarters, if not years, to evaluate. For now, the filings provide a rare, data-driven glimpse into how one of Wall Street’s most closely watched investors is positioning around the AI cycle – and which parts of the technology stack he believes will capture the largest share of the spoils.



