Billionaire Stanley Druckenmiller Sold Broadcom and Bought These Artificial Intelligence (AI) Giants Instead

Stanley Druckenmiller fully exited Broadcom and reallocated toward AI compute hyperscalers, buying sizable stakes in Amazon and Alphabet as hyperscalers gain more control over custom AI chips. The article argues both stocks trade at historically low forward P/E multiples—20.5x for Amazon and 16.4x for Alphabet—despite depressed free cash flow from heavy capital spending (Amazon FCF at -$7.6B over the last four quarters; Alphabet FCF at -$5.9B last quarter). With large contracted backlogs (Amazon $496B; Alphabet $514B in remaining performance obligations), it expects improved returns once new data centers come online, framing the setup as an opportunity despite near-term capex-driven cash burn.

Analysis

The market is mislabeling this as an AI beta trade; the real signal is a shift in bargaining power from chip vendors to the owners of demand, data, and deployment. If AMZN and GOOGL keep bringing more workloads onto self-directed silicon, the economics migrate up the stack: better pricing power, higher utilization, and more control over unit costs, while merchant suppliers lose some ability to dictate mix and margin. That makes AVGO and, to a lesser extent, MRVL more vulnerable on relative multiple expansion than headline AI spend would suggest.

Near term, the bullish setup for AMZN/GOOG is not earnings perfection but “capex now, operating leverage later.” The stock prices are still anchored to FCF optics, which can stay ugly for 1-2 quarters, but the next catalyst path is clear: capex commentary, cloud growth, and evidence that inference traffic is filling new capacity. If utilization ramps faster than depreciation rolls through, these names can re-rate before reported FCF recovers.

The contrarian risk is that the market may be too eager to buy the capex story before ROIC is visible. If AI workloads remain fragmented across third-party GPUs, or if hyperscaler buildout outruns demand, the current valuation support can evaporate quickly and the best names become value traps for several quarters. The thesis is falsified if management teams cut capex guidance, cloud growth slows, or custom silicon fails to show measurable share gains in the next two earnings cycles.

Key Decisions for Investors

• Long AMZN and/or GOOGL on 1-3 month weakness; use a staggered entry and hold for 6-12 months on the thesis that control of deployment and data gravity will convert capex into margin leverage. Risk/reward: asymmetric to the upside if utilization improves; cut if cloud growth or capex efficiency disappoints on the next two prints.
• Pair trade: long AMZN / short AVGO over the next 1-3 months. The mechanism is a transfer of negotiating power away from external silicon suppliers toward hyperscalers that can internalize more of the AI stack. Falsifier: AVGO re-accelerates on networking/custom ASIC mix or hyperscaler custom chip adoption stalls.
• If you want a cleaner expression, buy 3-6 month call spreads in GOOGL rather than outright stock. The options structure captures the rerating potential from depressed valuation while limiting damage if FCF optics stay negative for another quarter.
• Watch NVDA relative performance versus AMZN/GOOGL as a sentiment gauge for the AI supply chain. A persistent underperformance in NVDA versus hyperscalers would confirm the market is pricing in control-shift, not just AI demand growth.