I'd Buy Amazon Stock While It Sits 12% Below Its Record

Amazon shares are down about 12% from their $284.02 August record, but the article argues the pullback creates a buying opportunity as AWS revenue growth accelerated to 37% year over year in Q2, with operating income up 64% to $16.6 billion. AWS backlog reached roughly $496 billion, including Anthropic's more than $100 billion 10-year commitment, supporting the AI infrastructure build-out. The principal risk is capital intensity: trailing-12-month property and equipment spending rose to about $169 billion and free cash flow swung to a $7.6 billion outflow, while an FTC and 22-state lawsuit targets Amazon's advertising-auction practices.

Analysis

AMZN’s near-term debate is not AWS demand but capital intensity: the market will increasingly capitalize AI infrastructure as a recurring requirement rather than treat it as a temporary investment cycle. That makes quarterly capex guidance, useful-life assumptions, and incremental AWS operating margin more valuation-relevant than reported backlog. A sustained gap between AWS growth and capex growth would pressure FCF-based holders even if EPS continues to beat through depreciation timing.

The advertising litigation creates asymmetric risk because the remedy matters more than the headline fine. Forced auction redesign, advertiser restitution, or disclosure requirements could lower take rates and impair the high-margin profit pool that subsidizes retail logistics and AI investment. This also potentially benefits GOOG and META at the margin: agencies may shift incremental commerce-media budgets toward platforms with more transparent auction mechanics, particularly during the 1-3 month period in which legal discovery and remedy theories become clearer.

The non-obvious positive is that tighter AI infrastructure economics could reinforce hyperscaler concentration. Smaller cloud providers and enterprise-owned data centers face higher financing and utilization risk, while AMZN can fund underutilized capacity through retail cash generation and advertising profit. However, the Anthropic concentration dynamic is a two-sided risk: contracted demand supports utilization, but a major customer’s model economics or funding constraints would make the backlog less bankable than a diversified enterprise commitment.

Consensus is likely too focused on whether the recent pullback restores a headline earnings multiple. The re-rating catalyst is proof that incremental AI capex generates stable, cash-paying workload utilization; absent that evidence, AMZN can remain range-bound despite strong revenue. The thesis is falsified by AWS deceleration toward the low-20% range, declining AWS incremental margins, or another material upward revision to annual capex without corresponding backlog conversion and operating-income guidance.

Key Decisions for Investors

• Accumulate AMZN in tranches over the next 1-3 months rather than chase a single entry; size against a 12-18 month thesis that AWS utilization converts capex into FCF. Add only after the next earnings release confirms AWS growth above 30% and no material deterioration in incremental segment margin.
• Use a pair trade: long AMZN / short GOOG in modest size through the next two earnings cycles if AWS growth remains above 30%. AMZN has greater AI infrastructure operating leverage; exit if AWS growth falls below 25% or Google Cloud materially outgrows AWS for two consecutive quarters.
• Do not underwrite an AMZN long on ad revenue durability alone. Set a legal-event alert for any FTC remedy proposal involving auction-price methodology, advertiser reimbursement, or mandatory auction transparency; reduce exposure on a remedy that directly constrains ad monetization rather than a financial penalty.
• Watch NVDA as a read-through rather than a direct paired position: another broad hyperscaler capex uplift is supportive of near-term GPU demand, but a capex cut framed around lower-than-expected cloud utilization would be a negative second-order signal for both NVDA demand expectations and AMZN’s AI return profile.