Amazon After Q3: Can AWS Margins, Ad Growth and Prime Economics Turn Beats into Durable Profitability?

Amazon posted EPS of $1.95 on revenue of $180.17 billion, topping Street expectations on both metrics. Operating income landed at $17.4 billion, but that figure included a $2.5 billion Federal Trade Commission settlement related to Prime sign-ups and approximately $1.8 billion of severance charges tied to corporate job reductions. Adjusting for these items, underlying operating income was roughly $21.7 billion, revealing more margin expansion than the GAAP headline implies.

Segment dynamics leaned into Amazon’s mix-shift to higher-margin businesses. AWS revenue rose to $33 billion, up 20.2% year over year, while advertising revenue reached about $17.7 billion, ahead of consensus. Core online stores grew roughly 10%, reflecting the impact of Prime Day and operational gains in fulfillment and transportation.

Guidance kept the momentum intact. Management projected fourth-quarter net sales of $206–$213 billion (midpoint $209.5 billion) and operating income of $21–$26 billion. Notably, Amazon has a history of conservative guidance and outperformance, a pattern reinforced by this quarter’s upside relative to prior outlook. Capital expenditures are now expected to total approximately $125 billion in 2025, with management indicating a likely increase in 2026 as AI and core infrastructure buildouts accelerate.

AWS growth re-accelerated to about 20.2% year over year, and its operating income rose 9% to approximately $11.4 billion—accounting for roughly two-thirds of Amazon’s consolidated operating profit. Operating margins ran in the mid-30s (about 34.6%), a level that underscores the broader profitability thesis: if AWS can maintain mid-30s margins while scaling, it can carry group-level margins higher even as first-party retail remains structurally lower margin.

Underpinning demand is a reported backlog of around $200 billion and the CEO’s commentary that both AI and core infrastructure are energized. Amazon added over 3.8 gigawatts of capacity in the past 12 months and plans another gigawatt in the fourth quarter, with a multi-year plan to roughly double overall capacity by the end of 2027. The $11 billion Project Rainier—dedicated to Anthropic’s generative AI models—has gone live, and Amazon expects 1 million Trainium2 chips to be in use by the end of 2025.

The competitive backdrop remains intense. Microsoft’s Azure grew 40% and Google Cloud 34% in the same period, reflecting a market that’s expanding rapidly but is sensitive to price, performance, and power availability. Sustaining mid-30s margins while onboarding power-hungry AI training and inference workloads will require meticulous workload mix management, pricing discipline, and efficiency gains from custom silicon and networking. The near-term swing factor is whether AI capacity additions monetize quickly enough to keep margins resilient as the depreciation and energy bills start rising.

Advertising revenue came in at approximately $17.7 billion, ahead of expectations and reinforcing its role as a high-margin profit lever. The ads engine monetizes Amazon’s proprietary commerce data, closing the loop between product discovery and conversion. It also leverages Prime’s dense engagement—where frequent shoppers create intent-rich surface area for ad placements—turning what was once a cost center into a monetization channel.

The strategic interplay runs deeper than direct ad dollars. Ads help offset the margin drag of first-party retail by improving unit economics and funding fulfillment speed. They also strengthen the third-party marketplace by delivering traffic and conversion, seeding more selection and fees. Over time, incremental ad load within Prime Video and shoppable video integrations could extend this flywheel across entertainment and retail, while maintaining relevance through better targeting and measurement.

Risks are not trivial. Ads are exposed to macro advertising cycles, retail conversion trends, and competitive pressure from other platforms. But with consumer demand holding up and Amazon’s measurement advantages, the setup favors sustained share gains—particularly if Prime Video monetization scales without degrading user satisfaction.

Prime remains the connective tissue across Amazon’s platform: it drives frequency and conversion, densifies logistics routes, and anchors retention with bundled entertainment, shopping tools and perks. Evidence of this flywheel surfaced in the quarter through improved operating margin trajectories in North America and International when excluding special charges, alongside solid online stores growth that benefited from Prime Day.

The economic question is how much incremental pricing power and ad load can be layered onto Prime without introducing churn or diminishing satisfaction. While the subscription has proven resilient, any packaging tweaks, Prime Video ad load shifts, or regional price moves need to be sequenced with sensitivity to consumer income dynamics and competitive offers. The payoff is potentially higher ARPU and better per-order economics as engagement deepens.

Operationally, Prime’s contribution to logistics density continues to lower the cost to serve, which—combined with ad revenue and marketplace services—supports improving unit margins even when physical retail carries thinner gross margin. The path to durable profitability is not just about charging more; it’s about capturing more value from each Prime interaction and ensuring delivery speed and reliability keep consumers locked in.

Amazon’s plan to eliminate around 14,000 corporate roles is intended to flatten layers and move faster, not simply to chase near-term EPS optics. The company is reorienting its workforce mix for an AI-first operating model, with management emphasizing internal mobility and the cultural imperative of smaller, more accountable teams. Cost discipline in overhead and process efficiency can coexist with an elevated capex agenda—especially when those investments are concentrated in moats like cloud infrastructure and custom silicon.

Capex is the other side of the ledger—and it’s rising. With full-year 2025 capital expenditures now expected around $125 billion and indications that 2026 will be higher, free cash flow optics will be volatile. Investors should frame this spend as a capacity prepayment that must be matched by rapid backlog conversion and high utilization. The degree to which Amazon can monetize AI data center investments—via differentiated silicon, networking, and managed services—will determine whether returns on invested capital hold up.

The profit flywheel is clear: if AWS can maintain mid-30s margins and the ads business compounds, group-level operating margins can expand even if retail remains a headwind. The gating factor is the timing of returns on data center investments versus the depreciation and power costs that hit the income statement first. Managing power procurement, efficiency, and workload mix will be pivotal to translating capex into durable profitability.

AWS: Track growth relative to hyperscaler peers, backlog conversion velocity, and margin trends against rising energy and depreciation costs. Watch the workload mix between AI training, inference, and core infrastructure—and the adoption curve for Trainium2. Utilization and time-to-monetization for newly added capacity will be critical leading indicators.

Advertising: Monitor headline growth, CPC/CPM trends, and retail conversion signals around key seasonal windows. Evaluate Prime Video ad monetization for incremental ARPU, ad load tolerance, and impact on engagement. The balance between performance ad growth and brand dollars will inform durability.

Prime and Retail: Observe churn and engagement metrics following any packaging or pricing moves. Track shipping cost per unit and delivery speed improvements as a proxy for ongoing logistics density gains. In International, look for continued operating margin normalization as scale and process improvements take hold.

Cash, Capex, and Energy Strategy: Follow free cash flow yield relative to capex cadence, returns on AI data center spend, and any disclosures around power sourcing and efficiency. Given power constraints across the industry, long-dated energy contracts and on-site generation partnerships could become competitive differentiators.

Regulatory and One-offs: Stay alert to incremental competition or privacy-related costs that may distort reported operating income and complicate year-over-year comparisons. Normalizing for one-offs will remain essential to assessing true margin trends.