Why I Can't Stop Buying Amazon Stock

Amazon’s AWS Data services grew to $37.6B in Q1 2026 (+28% YoY, fastest pace in 15 quarters) with a $364B backlog (excluding the $100B+ Anthropic deal). The investment case highlights Trainium/Graviton revenue crossing a $20B annual run rate (+triple-digit growth) and consolidated operating margin of 13.1% (record high), alongside ROE of 24.3% and 35x interest coverage. The key downside is balance-sheet pressure: trailing-12M free cash flow fell ~95% to $1.2B while long-term debt rose to $119.1B from $65.6B; the author argues monetization should catch up as capacity is utilized.

Analysis

The market is likely still treating this as a capex story, but the more important mechanism is option value: Amazon is effectively pre-paying for future cloud and AI capacity, and the equity will re-rate only when investors believe that installed base is turning into durable margin, not just revenue. Near term, that means FCF can stay mechanically weak even if fundamentals are healthy, so the stock can drift on every cash-flow print before the operating leverage shows up.

Relative winners are the infrastructure picks-and-shovels around power, racks, networking, and custom silicon supply; the losers are cloud peers that look more disciplined on paper but may lose share if they cannot match contracted capacity or custom-chip economics. MSFT and GOOGL are the clearest relative shorts if the next two quarters show Amazon converting backlog faster than peers can catch up, but that pair only works if AWS growth remains above the market’s confidence threshold.

The key risk is that the market is overcounting signed demand and undercounting deployment friction: power availability, integration delays, and lower-than-assumed returns on each incremental data center can keep the spend elevated while margins plateau. The thesis is falsified if AWS growth decelerates materially for two consecutive quarters, if operating margin stops expanding, or if debt-funded capex forces management to slow the build in 2H26. In that scenario, the stock loses its 'pre-paid growth' narrative and trades back on cash conversion rather than revenue momentum.

Key Decisions for Investors

• Long AMZN on pullbacks over the next 2-6 weeks; target a 6-12 month hold for backlog-to-earnings conversion. Risk/reward is favorable if investors start valuing contracted AI capacity as deferred revenue-like visibility, but cut if FCF remains near-breakeven into 2H26.
• Pair trade: long AMZN / short MSFT over 3-6 months. Thesis is that Amazon has the cleanest path from capex to monetization; MSFT is more vulnerable if cloud growth slows while capital intensity stays high. Falsify on any meaningful AWS deceleration or Azure re-acceleration.
• Alternative relative value: long AMZN / short GOOGL for the same horizon. This expresses the view that market share and custom-silicon economics matter more than near-term cash discipline. Reduce if Alphabet’s capex begins to translate into visible enterprise AI wins.
• Basket long the infrastructure enablers on any broad AI-capex dip: VRT, ANET, ETN. These names monetize the buildout regardless of which hyperscaler wins share, but the trade is most sensitive to power and delivery lead times.