Amazon’s Quiet Chip Bet Is Becoming a Compounding Machine
Amazon’s cloud business just posted its fastest growth in nearly five years, and the smart money is paying attention. Goldman Sachs added 5.5 million shares of Micron and over 465,000 shares of Amazon in its latest 13F filing, while billionaire investors Stanley Druckenmiller, Peter Thiel, Seth Klarman, and David Tepper all built or expanded Amazon positions during the second quarter. Amazon is now the single most widely held stock among tracked billionaire portfolios — 62 of them, up from 59 the quarter before. That’s not a momentum trade; it’s a long-term conviction signal from investors who typically hold for years, not weeks.
The number driving that conviction: AWS revenue grew roughly 37% year-over-year in the second quarter, the fastest pace in 18 quarters, as AI workloads filled data center capacity faster than Wall Street modeled. Less discussed but arguably more important is what’s happening underneath the cloud number. Amazon’s custom silicon business — the Trainium and Graviton chip lines that reduce reliance on Nvidia and cut AI infrastructure costs — has already crossed a $25 billion annualized revenue run rate and is compounding at triple-digit percentage growth. That’s a vertically integrated moat few competitors can match: Amazon isn’t just renting out compute, it’s designing and selling the chips underneath it.
For patient investors, the real debate is about capital discipline, not growth. Amazon expects roughly $220 billion in capital expenditures this year, and free cash flow has already compressed under that spending. Management’s counterargument is that servers and networking hardware typically break even in under three years, and much of its AI capacity is pre-contracted for multi-year terms — a structural difference from speculative buildouts elsewhere in the sector. If that math holds, today’s margin pressure converts into a multi-year earnings and cash-flow acceleration once the current investment cycle matures. If AI demand cools before that payback window closes, the capex bill comes due with less to show for it.
Advertising and logistics automation add a second and third growth engine that rarely make headlines but continue to expand at a steady clip, giving Amazon more than one lever if cloud growth normalizes. So what for long-term investors: this isn’t a story about chasing an AI headline, it’s about whether a company with three compounding, cash-generative businesses can absorb a historically large capex cycle without permanently impairing returns on capital. The 13F data suggests some of the market’s most patient, valuation-conscious investors are betting it can.