Billionaire Ken Griffin Is Shifting His AI Bets. Here's What He's Buying and Selling.

Citadel’s AI tilt is clear: Ken Griffin sold 6.8M shares of Nvidia (cut exposure by 36%), trimmed Broadcom by ~11% (356k shares), and reduced Micron by 4M shares (down ~87%), while adding to Amazon and Microsoft. Amazon reported AWS fastest growth in 4+ years with quarterly AWS operating income rising to $16.6B (from $10.2B a year ago) and AI driving $25B annual run rates; Microsoft generated $19.6B in Q4 FY2026 free cash flow and remains free-cash-flow positive for FY2027. Since those late-July earnings reports, AMZN is up ~15% and MSFT up ~27%, reinforcing investor reward for demonstrated AI spending execution.

Analysis

This reads as a factor rotation from “AI spend as a story” to “AI spend with visible monetization.” AMZN and MSFT deserve a higher quality-of-earnings multiple because the market can now underwrite AI capex off internal cash generation rather than external financing, which lowers dilution/credit risk and supports a longer duration profile. That makes them the cleaner beneficiaries over the next 1-3 months if investors keep rewarding proof over promise.

The less obvious loser is not NVDA alone, but the basket of pure hardware beta that trades on accelerating shipment expectations. If hyperscaler capex remains elevated, the demand wave is still intact, but the mix may shift toward custom silicon, networking, and power infrastructure rather than incremental upside in GPU/memory multiples; that caps near-term upside for NVDA, AVGO, and MU unless they can show demand elasticity and pricing power into the next guide cycle. The key falsifier is any evidence that AWS/Azure growth is decelerating while capex stays high, which would force a sharper reset in cloud margin assumptions.

Contrarian view: the market may be too quick to call a winner-take-most regime in favor of the platforms and too slow to recognize that AI infrastructure is still in an early supply-constrained phase. In that setting, shorting the semis outright is risky because the next leg of demand can still surprise to the upside for several quarters. The better expression is relative value: own the names that can fund the buildout internally and hedge with the parts of the ecosystem most exposed to multiple compression if monetization slips.

Key Decisions for Investors

• Long AMZN on 3-5% pullbacks over the next 1-3 weeks; target 8-12% upside over 3 months if AWS growth and AI monetization remain intact. Falsifier: a meaningful deceleration in cloud growth or margin expansion.
• Long MSFT vs short NVDA as a 1-3 month pair trade; use MSFT's free-cash-flow resilience as the lower-beta AI winner while hedging outright AI exposure. Risk/reward improves if the market keeps rewarding monetization over shipment growth.
• Reduce tactical exposure to SMH / SOXX into the next hyperscaler capex commentary window; the best-case for semis is already well-owned, while any guide disappointment can compress multiples fast. Cover if NVDA/AVGO print upside revisions or memory pricing tightens further.
• If already long AI hardware, hedge with NVDA put spreads into the next cloud earnings cycle rather than selling stock outright; this keeps upside participation if capex stays hot while limiting downside if monetization cracks.