Amazon Crossed 50% Gross Margin. The Market Still Thinks It's a Retailer.

Amazon operates three reportable segments: North America, International, and AWS. That is the official structure. The more useful frame is: one capital-intensive retail and logistics network, and two high-margin businesses that piggyback on the infrastructure that retail built.

AWS is a cloud computing business generating approximately $107 to $115 billion in annualized revenue with operating margins above 38%. It competes with Microsoft Azure and Google Cloud. It wins on breadth of services, developer familiarity, and the fact that it was first. AWS market share has remained stable for over a decade, which is a stronger statement than it sounds in a market that was supposed to commoditize cloud infrastructure.

The advertising business is different. It does not have its own segment code. It sits inside North America and International revenue as online advertising services, and it is growing fast. Amazon's advertising revenue exceeded $56 billion in 2024 and grew meaningfully through 2025. The margin profile of advertising is close to pure software: once the platform infrastructure exists, incremental advertising revenue flows through at 60 to 70 percent margins.

Amazon's advertising competes differently from Google and Meta. The inventory is purchase-intent driven. A sponsored product listing appears when a consumer is actively trying to buy something. That is the highest-value advertising context in existence. It is not brand awareness. It is not passive scrolling. It is the moment of purchase, with a checkout button three taps away.

These two businesses, AWS and advertising, are why the gross margin is at 50% and heading higher.

The gross margin expansion from 42% in 2021 to 50.3% in 2025 is not random. It tracks directly with the growth of high-margin revenue relative to total revenue.

Revenue grew from $469.8 billion in 2021 to $716.9 billion in 2025, a 53% increase over four years. Gross profit grew faster: from $197.5 billion to $360.5 billion, an 83% increase. When gross profit grows twice as fast as revenue, that is a mix shift. High-margin segments are growing faster than the aggregate.

The operating leverage is even more dramatic. Operating income went from $24.9 billion in 2021 to $80 billion in 2025. Operating margin went from 5.3% to 11.2%. That is more than a doubling of the margin percentage in four years, on a base that grew by more than $247 billion in revenue. Very few companies demonstrate this kind of operating leverage at scale.

Net income tells the same story with more volatility. Amazon posted a loss of $2.7 billion in 2022, largely due to a Rivian investment writedown distorting the bottom line. Underlying operating performance never approached breakeven. By 2025, net income was $77.7 billion, up 31% from $59.2 billion in 2024.

Earnings per share has beaten consensus estimates for six consecutive quarters, with surprise percentages ranging from 16.9% to 26.6%. That is not one good quarter. That is a structural pattern of management guiding conservatively on a business that is consistently outperforming. The analyst consensus, with 47 strong buy ratings and zero sells, prices in continued execution. So far the execution has warranted it.

Amazon's advertising business is the least discussed major growth driver in U.S. equities. A business generating more than $60 billion in annual revenue, growing at double-digit rates, with margins that rival the best software companies on the planet, and it does not have its own segment line item.

The reason is structural: advertising revenue is reported inside the North America and International retail segments. Investors focused on AWS treat the retail numbers as undifferentiated. They are not. The advertising portion behaves entirely differently from product sales: no inventory, no fulfillment cost, near-zero marginal cost per incremental dollar. The reporting structure obscures the economics.

Analysts specifically flagged Amazon's advertising revenue momentum as a potential upside driver going into 2026 reporting. That framing understates the point. The advertising business is not a potential driver. It is an existing structural driver that is already compressing the gap between reported financials and what a sum-of-parts analysis would suggest the business is worth.

The competitive dynamic is shifting in Amazon's favor. Google's search advertising faces structural questions about AI-generated answers displacing traditional search queries. Meta faces ongoing regulatory pressure in Europe. Neither headwind applies to Amazon's purchase-intent advertising inventory. When a consumer searches for a product on Amazon, the advertiser is paying for attention from someone who intends to buy. That context does not erode with AI overviews.

Sponsored product and brand listings have become a mandatory cost of doing business for consumer goods companies. Amazon controls the discovery layer of the largest product catalog in the world. Sellers who want to be found pay. That dynamic does not reverse easily, and pricing power compounds over time as the catalog grows and the algorithm improves.

AWS generated approximately $107 to $115 billion in revenue in 2025, with operating margins consistently above 38%. The segment is one of the highest-margin large-scale infrastructure businesses in the world. That is the starting point for any serious Amazon analysis.

The key question for AWS is not whether it is profitable. It is whether the growth rate is sustainable. Cloud infrastructure spending is tied to enterprise software adoption, AI workload migration, and digital transformation broadly. The AI infrastructure buildout has added a new demand layer that did not exist three years ago. AI training and inference workloads are computationally intensive, require specialized infrastructure, and are increasingly being run on cloud platforms rather than on-premise.

Amazon's AI strategy within AWS centers on Bedrock, the managed service for accessing foundation models from multiple providers, and its custom Trainium and Inferentia silicon designed to run AI workloads more efficiently than standard GPU configurations. The Kuiper satellite internet program expanded its commercial partnerships on March 31 with a deal to provide Delta Air Lines with in-flight connectivity, adding another access layer for AWS cloud services in environments where terrestrial internet infrastructure is absent or limited.

The breadth argument for AWS is genuine. AWS offers more than 200 distinct cloud services. The competitive advantage of breadth is not just having the services. It is that migrating away from AWS requires rebuilding integrations across hundreds of service touchpoints. Switching costs are real, measurable, and structurally defended.

Microsoft Azure has taken some share in recent years, driven by Microsoft 365 integration and the Copilot AI wave. But the cloud market is expanding fast enough that both players growing simultaneously is the base case, not a zero-sum competition. Amazon's position is defensible.

Amazon spent $131.8 billion in capital expenditures in 2025. That is up from $83 billion in 2024. Free cash flow compressed to $7.7 billion as a result, down from $32.9 billion the prior year.

This is the number that unsettles some investors. A $2.1 trillion market cap company generating $7.7 billion in FCF screens as expensive on any price-to-FCF metric. That is the wrong frame. Operating cash flow was $139.5 billion in 2025. Amazon chose to reinvest $131.8 billion of that in infrastructure. The FCF compression is a capital allocation decision, not a deterioration of the underlying business. The business itself is generating cash at an accelerating rate.

The capex is going into data centers, AI compute chips, satellite infrastructure, and last-mile logistics. These are not speculative bets on unknown markets. AWS data centers serve existing enterprise customers under multi-year contracts. The demand for AI compute is structurally increasing. The capex is spending into demonstrated demand, not ahead of it.

Amazon does not pay a dividend and has not conducted meaningful share buybacks. Shares outstanding have grown slightly from dilution, with stock-based compensation expense of $19.5 billion in 2025. The share count has grown from 10.26 billion in 2020 to 10.73 billion in 2025, a 4.6% increase over five years. By the standards of high-SBC technology companies, this is modest.

The implicit management thesis: deploying capital into AI infrastructure now generates superior returns over the next decade relative to returning cash to shareholders. Based on the AWS margin trajectory and the structural position of the business, that bet has earned some credibility.

At a $2.16 trillion market cap and $165.3 billion in EBITDA, Amazon trades at approximately 13.5x EV/EBITDA on a trailing basis. That multiple is strikingly low relative to the quality of the business.

Microsoft trades at 20 to 25x EV/EBITDA. Alphabet at 12 to 16x. Meta at 14 to 18x. Amazon, with higher EBITDA growth of 33.5% in 2025, broader business diversification, and two structural high-margin drivers still in growth mode, trades at the low end of the large-cap technology peer group.

The trailing P/E of 28x appears more demanding. But earnings have grown 31% year-over-year, and the Q1 2026 consensus estimate of $1.73 per share implies annual EPS approaching $7.50 to $8.00 if the quarterly trajectory holds. On a forward basis, the multiple compresses materially. The six-quarter EPS beat pattern, with surprise percentages ranging from 17% to 26.6%, suggests that either management guides conservatively by design or analyst models consistently underestimate operating leverage. Both have the same valuation implication: the consensus estimate is likely too low again.

Market sentiment for Amazon has averaged above 0.70 on a normalized scale over the past 30 days, with readings ranging from 0.39 to 0.97. The consistently positive sentiment is tracking genuine business developments: advertising momentum, AWS AI workload growth, and new commercial partnerships. Sentiment divergences are where analysis gets interesting. Here, positive sentiment and an undervalued fundamental profile point in the same direction. That is not the most analytically complex setup, but it is a useful one.