Amazon Hasn't Repurchased a Single Share in 4 Years. That's Exactly Why You Should

BofA argues hyperscalers are shifting hundreds of billions from buybacks to AI infrastructure, pressuring free cash flow near term while semiconductor cash generation benefits from rising AI chip demand. Amazon highlights the divergence: it has not repurchased shares since Q2 2022 and is instead funding AWS AI data centers plus custom Trainium/Inferentia chips and AI-focused infrastructure. The article frames the tradeoff as potentially higher long-run returns, noting Amazon trades ~29x earnings (near historical lows) with analysts projecting ~22% average annual earnings growth over the next five years, despite execution risk from elevated capex.

Analysis

The market is still treating this as a capital-return story, but the more important read-through is that hyperscalers are substituting financial engineering with internal investment at a moment when their incremental ROIC likely still clears cost of capital. That is constructive for AMZN over a 6-18 month horizon because its spend has the clearest path to monetization through AWS attach, ads, logistics efficiency, and custom silicon leverage. Near term, the main loser is not AMZN shareholders per se but any investor base anchored to buyback-driven EPS support; that cohort can keep the stock range-bound until FCF conversion visibly improves.

The second-order winners are the picks-and-shovels suppliers to the AI buildout: semis, networking, power, and data-center infrastructure. But the trade is becoming less about raw chip demand and more about which vendors can keep pricing power once hyperscalers internalize more of the stack. That creates a subtle headwind for standalone accelerator economics over time if Trainium/Inferentia-style substitution gains traction, even as overall AI capex stays strong.

Risk is execution, not narrative. Over the next 1-3 months, the stock will trade on capex guidance and whether revenue growth inflects fast enough to offset FCF compression; if spend rises without AWS or ads reaccelerating, multiple support can evaporate. The contrarian view is that the market may be over-penalizing the absence of buybacks: if management is right on capital allocation, AMZN’s valuation can re-rate before the cash flow does, but that requires evidence on the next two earnings prints, not just optimism.

Key Decisions for Investors

• Long AMZN on 2-3% pullbacks; 6-12 month horizon targeting a re-rating if AWS/ads growth absorbs capex and FCF inflects. Falsifier: another quarter of rising capex with no margin or revenue leverage.
• Pair trade: long AMZN / short QQQ or XLY for the next 1-3 quarters if the market stays fixated on buybacks; thesis is AMZN-specific operating leverage against broader mega-cap multiple compression.
• Alert on semis via SMH or SOXX rather than single-name chasing; maintain longs only while hyperscaler capex remains upwardly revised. Falsifier: a capex guide reset or enterprise AI spend delay.
• Watch NVDA/AVGO for any sign that hyperscaler custom silicon is displacing merchant accelerators faster than expected; if that substitution evidence emerges, rotate from pure compute beta into networking/power beneficiaries.