Amazon is increasingly becoming a different company from the one investors owned a decade ago. E-commerce remains the foundation of its consumer ecosystem, but a growing portion of Amazon's incremental value is being generated by AWS, artificial intelligence, custom silicon and advertising. The result is a business where the fastest-growing and highest-margin operations are becoming increasingly important to consolidated earnings.

The central debate surrounding AMZN today is therefore not whether Amazon has attractive businesses. It is whether the company can earn sufficient returns on the extraordinary amount of capital it is deploying into AI infrastructure. Amazon expects approximately $220 billion of capital expenditures in 2026, while trailing free cash flow has temporarily turned negative. Yet AWS is simultaneously experiencing its fastest growth in 18 quarters, capacity remains constrained, and management says it has clear visibility into attractive financial returns from the infrastructure being built.

The Amazon investment thesis rests on four major businesses working together: AWS, AI infrastructure, advertising and the retail/Prime ecosystem. AWS provides the largest source of operating profit and is accelerating as AI workloads move into production. Advertising monetizes Amazon's enormous pool of purchase-intent data at very high margins. Retail continues improving efficiency through automation, regional fulfillment and faster delivery, while Prime increases customer retention and spending across the ecosystem.

The most important emerging component is AI. Amazon isn't merely renting NVIDIA GPUs through AWS. It is building its own Trainium and Graviton chips, expanding Bedrock, constructing enormous data centers and investing deeply in Anthropic. If those investments generate attractive returns, Amazon could become one of the largest beneficiaries of the AI infrastructure cycle while simultaneously reducing its dependence on third-party semiconductor suppliers.

Amazon's Q2 2026 results showed meaningful acceleration across the company. Revenue reached $200.6 billion, increasing 20% year over year, while operating income increased 43% to $27.5 billion. North America generated $116.2 billion of revenue, International generated $42.2 billion and AWS generated another $42.2 billion.

The headline GAAP net income figure requires an important adjustment. Amazon reported $62.6 billion of net income, but that included approximately $53.4 billion of non-operating pretax income, primarily associated with the increased value of Amazon's Anthropic investment. Investors therefore shouldn't treat the resulting headline EPS as representative of Amazon's underlying earnings power.

AWS remains the centerpiece of the investment thesis. Q2 revenue reached $42.2 billion, up approximately 37% year over year and representing the fastest AWS growth rate in 18 quarters. That puts AWS at approximately a $169 billion annualized revenue run rate.

Even more important is profitability. AWS generated approximately $16.6 billion of operating income, compared with $10.2 billion one year earlier, implying an operating margin around 39.4%. AWS therefore represented only about 21% of Amazon's revenue but approximately 60% of consolidated operating income.

This is why AWS growth matters disproportionately to AMZN's valuation. Every incremental dollar of AWS revenue carries substantially greater earnings power than a typical dollar of Amazon retail revenue.

Amazon isn't currently struggling to find customers for its AI infrastructure. Its larger problem is building capacity quickly enough to satisfy them. AWS's contracted backlog has expanded dramatically, with significant portions of future capacity already committed.

Management has repeatedly indicated that AI demand continues exceeding available capacity and that much of the infrastructure currently under construction has customers waiting for it. This distinction matters enormously for the capex debate: spending $220 billion speculatively would be concerning, while spending aggressively against contracted or highly visible demand presents a substantially different risk profile. Amazon still needs to prove the returns, but the current evidence suggests the infrastructure buildout is being pulled by customer demand rather than pushed solely by management ambition.

The largest financial issue surrounding Amazon is its planned approximately $220 billion of 2026 capital expenditures, the majority of which supports AWS and artificial intelligence infrastructure. Q2 cash capex alone reached roughly $53 billion.

That spending is enormous even by Amazon's standards, but management argues that the economics remain attractive. Data centers have long useful lives, while servers and networking equipment have shorter replacement cycles and can generate returns relatively quickly once deployed. Management has said it has clear visibility into strong returns at current spending levels.

This creates the key question for investors: Is Amazon temporarily suppressing free cash flow to create a dramatically larger AWS earnings base, or is the company entering a structurally lower-return capital-intensive era? My base case assumes the former, but this is the most important variable to monitor.

Amazon generated $161.4 billion of trailing operating cash flow, up 33%, demonstrating that the underlying businesses remain extraordinarily cash generative. Yet trailing free cash flow fell to an outflow of $7.6 billion, compared with positive $18.2 billion one year earlier, primarily because purchases of property and equipment increased dramatically to support AI infrastructure.

That divergence between operating cash flow and free cash flow is critical. The business itself isn't suddenly deteriorating; Amazon is deliberately reinvesting an enormous portion of internally generated cash into infrastructure. If AWS growth and AI demand remain strong, free cash flow could recover significantly once the rate of infrastructure expansion eventually moderates. If demand disappoints, however, today's capex becomes a major source of value destruction.

Amazon's AI strategy extends considerably beyond renting NVIDIA GPUs. The company is building an integrated stack consisting of AWS infrastructure, Trainium accelerators, Graviton CPUs, Bedrock model access, custom networking, proprietary software and massive data-center capacity.

AWS's AI business has already surpassed a $25 billion annual revenue run rate and is growing at triple-digit rates. Amazon's chips business has separately surpassed a $25 billion annual revenue run rate and is also growing at triple-digit rates.

This matters because Amazon has an opportunity to capture economics from multiple layers of the AI infrastructure stack instead of merely reselling compute produced by other semiconductor companies.

Trainium may ultimately become one of Amazon's most strategically important assets. NVIDIA remains deeply embedded within AWS, and management has explicitly said customers will continue using NVIDIA hardware for the foreseeable future. But Amazon increasingly wants customers to have a lower-cost alternative optimized for training and inference.

Trainium gives Amazon greater control over cost, power efficiency, supply and pricing while potentially improving AWS margins. Graviton performs a similar function for general-purpose computing and is already used by 98% of AWS's top 1,000 EC2 customers. Amazon said Graviton revenue commitments increased nearly threefold quarter over quarter.

The likely future isn't AWS abandoning NVIDIA. It is AWS offering NVIDIA for customers demanding maximum flexibility and ecosystem compatibility while increasingly using proprietary silicon where Amazon can deliver superior economics.

Amazon's relationship with Anthropic has evolved far beyond a traditional equity investment. Amazon announced another $5 billion investment and could invest up to an additional $20 billion if certain commercial milestones are reached. More importantly, Anthropic has committed to securing up to 5 gigawatts of AWS compute capacity and spending more than $100 billion on AWS technologies over ten years.

Project Rainier already uses more than one million Trainium2 chips to train and serve Claude. Over 100,000 AWS customers use Claude models, making Anthropic simultaneously an investment, a major customer, a Trainium validation partner and one of the most important model providers available through AWS.

That relationship creates substantial strategic value, although investors should separate the mark-to-market value of Amazon's Anthropic stake from recurring operating earnings.

Amazon Bedrock is AWS's managed platform for accessing and deploying foundation models. Rather than forcing customers into one proprietary model ecosystem, Amazon offers models from multiple providers and lets enterprises deploy them using AWS infrastructure, security and data.

Hundreds of thousands of customers now use Bedrock, and Amazon reported that more customers were added during the most recent six months than during the platform's first two years combined. Customer spending on Bedrock in Q2 also exceeded all prior quarters combined.

This multi-model approach is strategically important. Amazon doesn't necessarily need Claude, OpenAI or another model provider to dominate. If enterprise AI workloads run through AWS regardless of which model wins, Amazon can participate across the ecosystem.

Advertising has become one of Amazon's most valuable businesses. Q2 advertising revenue approached $20 billion and grew approximately 26% year over year, putting the business near an $80 billion annualized revenue run rate.

Amazon possesses a significant advantage because users frequently arrive with direct purchasing intent. Advertisers can target customers close to the moment of purchase and measure whether advertising translates into actual sales. That closed-loop attribution is extremely valuable and differentiates Amazon from traditional digital advertising platforms.

Advertising also carries substantially better economics than first-party retail, meaning continued ad growth can steadily increase consolidated margins even if Amazon's overall revenue mix changes only gradually.

Retail remains Amazon's largest business by revenue and the foundation of its consumer ecosystem. North American sales increased 16% to $116.2 billion, while operating income increased to approximately $9.1 billion. International revenue increased 15% to $42.2 billion and produced approximately $1.7 billion of operating income.

The key change is that Amazon has spent years making fulfillment more efficient through regionalized inventory, automation and faster delivery. More than 40% more items were delivered same-day or overnight during the first half of 2026, while grocery and everyday essentials grew faster than the broader retail business.

Prime strengthens this ecosystem by increasing purchasing frequency, retention and engagement across shopping, entertainment and other Amazon services. Retail therefore matters not only for direct profits but because it feeds Amazon's logistics, advertising and subscription businesses simultaneously.

Amazon's logistics network remains one of the company's strongest competitive advantages. Years of investment in fulfillment centers, delivery stations, transportation capacity, robotics and inventory placement have created infrastructure few competitors could replicate economically.

AI and robotics should continue improving this network. Amazon recently introduced a new generation of its Proteus autonomous robot capable of moving loads of up to 1,300 pounds and responding to conversational instructions. Automation can reduce fulfillment costs, improve delivery speeds and increase throughput without requiring labor expenses to rise proportionally with volume.

This is an underappreciated source of potential margin expansion because even small efficiency gains become financially significant across Amazon's enormous retail operation.

Amazon's moat comes from several reinforcing advantages rather than one product. The most important are:

The businesses reinforce one another. Retail creates customer traffic. Traffic improves advertising. Prime increases customer retention. AWS produces high-margin earnings. Those earnings help fund infrastructure. Infrastructure strengthens AWS and retail. That flywheel is extremely difficult for competitors to reproduce in its entirety.

Andy Jassy's tenure has increasingly focused on improving Amazon's cost discipline while simultaneously investing aggressively where management sees large long-term opportunities. The company has reduced unnecessary expenses, improved fulfillment economics and expanded margins while directing extraordinary amounts of capital toward AWS and AI.

That strategy creates an unusual combination: operational discipline in mature businesses alongside enormous investment in emerging infrastructure. The critical test of management over the next several years will be whether the roughly $220 billion capex program produces sufficient incremental AWS revenue and operating income.

If it does, the current spending cycle could resemble Amazon's earlier fulfillment and AWS investments—periods where free cash flow temporarily looked weak because the company was building infrastructure that later became extraordinarily valuable.

Amazon maintains a substantial liquidity position and generates enormous operating cash flow, giving it considerably more flexibility than most companies attempting infrastructure programs of comparable scale.

The key issue isn't solvency or liquidity; it's capital efficiency. Amazon can afford to spend $200+ billion annually. The question is whether shareholders should want it to. Investors should therefore focus less on the absolute capex number and more on incremental AWS revenue, operating margins, contracted backlog, utilization and eventual free cash flow generated by those investments.

Amazon continues diversifying the technology underlying AWS. In September, Qualcomm announced a multi-generational collaboration with Amazon involving customized silicon for large-scale AI inference and advanced optical connectivity for AWS data centers.

This fits Amazon's broader strategy of controlling more of its infrastructure economics. AWS can simultaneously deploy NVIDIA accelerators, expand Trainium, use Graviton CPUs and collaborate with semiconductor partners on specialized components. Over time, that architecture could reduce AWS's cost per unit of compute and improve its ability to compete on price without sacrificing margins.

The biggest catalyst is continued AWS acceleration. If AWS sustains roughly 30%+ growth while maintaining strong operating margins, Amazon's consolidated earnings can continue growing substantially faster than overall revenue.

Other catalysts include increasing Trainium adoption, Anthropic's enormous AWS commitments, Bedrock growth, advertising remaining above 20%, continued retail margin expansion and eventual normalization of capex growth. The last point could become particularly powerful: if operating cash flow continues expanding while infrastructure spending eventually stabilizes, reported free cash flow could rebound rapidly.

The upcoming earnings cycles therefore matter less for whether Amazon beats quarterly retail revenue by a few percentage points and more for whether AWS growth, backlog, AI capacity utilization and capex returns continue validating the investment cycle.

The largest risk is that AI infrastructure spending fails to generate adequate returns. Amazon is committing enormous amounts of capital based on the assumption that demand for AI training and inference will continue expanding for years.

Other important risks include competition from Microsoft Azure and Google Cloud, custom silicon execution, dependence on NVIDIA for portions of high-end AI infrastructure, regulatory pressure, retail competition, wage and logistics costs, weakening consumer spending and potential margin pressure if AWS pricing becomes more competitive.

Anthropic also creates concentration risk. The relationship is strategically attractive, but Amazon has substantial financial and infrastructure exposure to one of the largest AI model developers. A major deterioration in Anthropic's competitive position could affect both the investment value and expected AWS demand associated with the partnership.

I would become materially less constructive if several developments occurred together:

The most important relationship is AWS revenue and operating-income growth versus AI capex growth. As long as infrastructure investment continues producing accelerating cloud revenue and earnings, the thesis remains healthy. If capex remains enormous while AWS growth rolls over, the investment case changes substantially.

Amazon is difficult to value using one traditional multiple because its businesses have radically different economics. Retail, AWS and advertising shouldn't logically receive the same valuation multiple, while headline GAAP earnings are currently distorted by the mark-to-market appreciation of Amazon's Anthropic investment.

A sum-of-the-parts framework therefore provides useful context. Using approximate annualized business levels rather than pretending these values are precise:

This isn't intended as a mechanical target-price calculation. It demonstrates why Amazon can support a multi-trillion-dollar valuation even though traditional retail represents much of consolidated revenue. AWS and advertising carry dramatically better margins and potentially deserve much higher valuation multiples.

The bear scenario assumes AWS growth decelerates toward the low-to-mid 20% range, Amazon continues spending approximately $200 billion or more annually on infrastructure, free cash flow remains weak, advertising growth moderates and the market begins questioning whether AI infrastructure can earn sufficient returns. Under that scenario, the market would likely assign a lower multiple to normalized earnings and cash flow.

My base case assumes AWS remains roughly a 30%+ growth business, advertising remains around or above 20%, retail margins continue gradually improving and Trainium gains adoption alongside NVIDIA infrastructure. Capex remains elevated but increasingly produces revenue and operating income, allowing investors to look through temporarily depressed free cash flow.

The bull case requires AWS to sustain approximately 35% or better growth, Trainium to become a meaningful alternative AI-compute platform, Anthropic and other customers to consume enormous AWS capacity, advertising to maintain mid-20% growth and retail margins to expand further. If capex subsequently stabilizes while operating cash flow continues climbing, Amazon could experience a significant free-cash-flow inflection that warrants a materially higher valuation.

Amazon currently offers an unusual combination of mature businesses capable of producing enormous cash flows and emerging businesses still growing at venture-like rates. AWS, advertising and custom silicon increasingly drive incremental economics, while retail provides the enormous customer ecosystem underlying Prime, logistics and advertising.

The investment case therefore comes down to whether Amazon can successfully convert its unprecedented AI infrastructure spending into sustainable high-margin revenue. So far, AWS's acceleration, backlog, capacity constraints and AI adoption suggest the spending is producing tangible demand. The next several quarters need to continue confirming that relationship.

The $320 target assumes AWS remains approximately a 30%+ growth business, advertising continues compounding above 20%, retail margins gradually improve and Amazon's massive AI infrastructure investment begins translating into proportionately larger revenue and operating income. It also assumes the market eventually looks through temporarily depressed free cash flow as the current capex cycle begins demonstrating attractive returns.

The central reason for the BUY rating is that Amazon's highest-quality businesses are becoming increasingly important to consolidated profitability. AWS represents roughly one-fifth of revenue but around 60% of operating income, advertising is approaching an $80 billion annualized business, and Amazon's AI and chips businesses have each surpassed $25 billion annualized revenue run rates.

The principal risk is straightforward: Amazon is spending approximately $220 billion this year largely because management believes AI infrastructure demand will justify it. If AWS growth remains around 30–40%, backlog converts into revenue and Trainium gains adoption, today's capex could create an enormous future earnings base. If AWS growth rolls over while spending remains elevated, free cash flow and valuation would come under significant pressure.

For now, the evidence favors the constructive scenario. AWS just produced its fastest growth in 18 quarters, operating income is growing substantially faster than consolidated revenue, AI demand remains capacity constrained, and Anthropic alone has committed more than $100 billion to AWS technologies over the coming decade.

Newsletter Verdict: BUY — $320 12-month target.

This rating and valuation are an equity-research framework based on the assumptions above, not individualized investment advice.