I have owned, watched and analyzed enough technology stocks to recognize the familiar stages of an artificial intelligence investment story.
First comes amazement.
Then comes excitement.
Next comes a corporate presentation containing the word “AI” so many times that I begin to wonder whether the accounting department has been replaced by a chatbot.
Finally, investors ask the only question that matters: Where is the money?
Amazon is moving beyond the presentation stage.
AWS is not merely experimenting with artificial intelligence or adding a cheerful assistant to an existing product. Amazon is spending extraordinary amounts of money to build the infrastructure, chips, models and software that it believes will power the next generation of computing.
The scale is breathtaking. It is also mildly terrifying.
Amazon expects to invest approximately $200 billion in capital expenditures during 2026. That is not a typo caused by an analyst falling asleep on the zero key. The company is spending an amount larger than the annual economic output of many countries, with much of that investment directed toward technology infrastructure and AWS.
The optimistic interpretation is that Amazon is building the toll roads, power stations and industrial machinery of the AI economy.
The cautious interpretation is that it is spending $200 billion because every technology giant currently believes there can never be too many data centers.
I believe AWS can drive Amazon’s next leg higher. The latest numbers give me more confidence than I had a year ago. But I do not believe investors should ignore the cost, execution risk or increasingly aggressive competition surrounding this opportunity.
At approximately $258.63 per share as of August 21, 2026, Amazon is no longer a neglected bargain hiding beneath widespread pessimism. Its market capitalization is roughly $2.82 trillion. Expectations are substantial, and the company now needs to prove that its historic investment cycle will generate equally historic returns. Market data currently shows a trailing price-to-earnings ratio near 21, although that figure is unusually flattering because Amazon’s recent earnings include a massive investment-related gain. Current AMZN market data
My view is simple: I consider Amazon a long-term buy, but not because someone placed the letters “AI” beside the ticker.
I consider it a buy because AWS growth is accelerating, AWS margins are strong, Amazon’s custom chips could improve its economics, and the company possesses an unusually broad collection of ways to turn AI infrastructure into recurring revenue.
The opportunity is real.
So is the bill.
My Amazon Investment Thesis in One Sentence
If I had to summarize my thesis without using 47 slides and an animated diagram of a data center, I would say this:
AWS gives Amazon a credible chance to own several profitable layers of the AI economy at once—provided today’s enormous capital spending becomes tomorrow’s durable revenue rather than tomorrow’s very expensive surplus capacity.
Amazon does not need to develop the single most powerful AI model in the world to win.
It can provide the computing infrastructure on which models are trained. It can sell inference capacity when those models are used. It can offer custom chips that reduce costs. It can distribute third-party models through Amazon Bedrock. It can sell tools that help businesses build, secure and operate AI agents. It can weave AI into retail, advertising, logistics, Alexa, healthcare and workplace software.
In other words, Amazon does not need to predict which model developer becomes the ultimate champion. AWS can sell equipment to nearly everyone participating in the competition.
That is a position I find attractive.
During a gold rush, I generally prefer the company selling reliable machinery to the person insisting he has found the one enchanted hill.
AWS Is Accelerating Again
The most encouraging part of Amazon’s latest report was not that AWS grew.
AWS has been growing for years.
The important development was the speed of the acceleration.
AWS generated $42.2 billion in second-quarter 2026 revenue, up 37% from $30.9 billion a year earlier. Amazon described that as AWS’s fastest growth in 18 quarters and an annualized revenue run rate of approximately $169 billion.
That is not the growth rate of a small business benefiting from easy comparisons. AWS is already enormous. Adding 37% growth to a quarterly revenue base exceeding $40 billion is the corporate equivalent of watching an aircraft carrier accelerate like a sports car.
Growth had already strengthened during the previous quarters. AWS revenue rose 20% in the third quarter of 2025, 24% in the fourth quarter and 28% in the first quarter of 2026. It then reached 37% in the second quarter.
That progression matters to me:
| Period | AWS revenue | Year-over-year growth |
|---|---|---|
| Q3 2025 | $33.0 billion | 20% |
| Q4 2025 | $35.6 billion | 24% |
| Q1 2026 | $37.6 billion | 28% |
| Q2 2026 | $42.2 billion | 37% |
The acceleration suggests that AI demand is moving from experimentation toward meaningful deployment.
Companies spent the early portion of the generative AI boom testing models, launching pilots and issuing press releases about strategic transformation. Some of those projects were probably useful. Others may have been PowerPoint presentations wearing safety goggles.
Now, more workloads appear to be reaching production.
Training large models consumes immense computing resources, but inference may become the larger and more durable opportunity. Inference happens whenever a trained model processes a request, generates an answer, analyzes an image, writes code or allows an AI agent to perform a task.
Training creates the intelligence.
Inference pays the recurring utility bill.
If AI becomes embedded in customer service, software development, medical research, advertising, logistics and everyday office work, inference demand could expand for years.
AWS is positioned to collect a portion of that spending every time businesses use the underlying infrastructure.
According to Amazon’s second-quarter release, its AWS AI business exceeded a $25 billion annual revenue run rate and was growing at a triple-digit percentage. Its chips business also surpassed a $25 billion annual run rate, again with triple-digit growth. Those figures indicate that AI has become a significant commercial business rather than a distant promise. Amazon’s second-quarter 2026 results
AWS Is Amazon’s Profit Engine
Revenue growth is delightful, but I become much more attentive when growth arrives with substantial operating income.
AWS produced $16.6 billion in operating income during the second quarter, up from $10.2 billion a year earlier. Its operating margin was approximately 39%.
To put that contribution in perspective, AWS accounted for roughly 21% of Amazon’s quarterly revenue but more than 60% of its operating income.
That is the part of the story I do not want investors to overlook.
Amazon’s retail operations generate staggering sales, but retail is capital intensive, logistically complicated and traditionally lower margin. AWS provides the profit engine that can fund investment across the broader company.
In the second quarter:
| Segment | Revenue | Operating income |
|---|---|---|
| North America | $116.2 billion | $9.1 billion |
| International | $42.2 billion | $1.7 billion |
| AWS | $42.2 billion | $16.6 billion |
AWS generated exactly as much revenue as Amazon’s International segment, yet it produced nearly ten times as much operating income.
That is why I regard AWS as more than one component of Amazon. It is the economic center of gravity.
If AWS continues growing faster than the company as a whole while maintaining strong margins, Amazon’s consolidated earnings can rise faster than its revenue. The mix shifts toward the more profitable business.
That is the mechanism through which AWS could drive the stock’s next move.
Investors do not need Amazon’s online stores to suddenly resemble a luxury software company. They need AWS to become a larger percentage of the earnings stream while retail, advertising and logistics continue improving around it.
Trainium May Be the Most Important Part of the Story
Nvidia remains the dominant provider of AI accelerators, and I am not interested in pretending otherwise. Its hardware and software ecosystem created an extraordinary competitive position.
But large cloud providers have powerful incentives to reduce their dependence on any single chip supplier.
Amazon’s answer is Trainium.
Trainium is Amazon’s custom accelerator designed for training and running AI models. Inferentia serves inference workloads, while Graviton handles more general computing tasks. Together with Amazon’s Nitro system, these chips give AWS greater control over performance, availability and cost.
This matters for two reasons.
First, customers want lower prices.
AI computing is expensive. If Amazon can provide acceptable or superior performance at a lower total cost, customers have an economic reason to adopt Trainium. They do not need to become emotionally invested in Amazon’s silicon. Corporate loyalty tends to develop rapidly when the finance department discovers meaningful savings.
Second, custom chips could improve AWS margins.
When AWS relies on third-party accelerators, part of the economics goes to the chip supplier. When Amazon designs more of the stack itself, it can retain more value.
Amazon has suggested that Trainium could eventually save it tens of billions of dollars in annual capital expenditures and provide several hundred basis points of operating-margin advantage for inference compared with depending entirely on outside chips. That is management’s projection, not a guaranteed outcome, but the potential is enormous. Amazon’s 2025 annual report discusses its custom-chip economics and AI investment cycle
The company is also reporting serious customer commitments. Amazon says Anthropic and OpenAI have made multi-year, multi-gigawatt Trainium commitments, alongside adoption from startups and larger enterprises.
These commitments are important because data-center construction requires money years before the revenue arrives. Amazon must acquire land, secure electricity, construct facilities, install networking equipment and deploy chips. It cannot wait for a customer to click “buy” and then build a data center during the delivery window.
The infrastructure has to exist first.
Customer commitments reduce—but do not eliminate—the risk that Amazon is building capacity nobody ultimately needs.
Bedrock Gives AWS a Model-Neutral Advantage
One of Amazon’s more intelligent AI decisions was not forcing customers into a single model.
Amazon Bedrock allows businesses to access and use multiple foundation models through AWS. In its second-quarter update, Amazon said Bedrock offered models from providers including Anthropic, OpenAI, Google DeepMind and others.
This resembles a well-stocked model supermarket.
One customer may prefer an Anthropic model for a particular agentic workflow. Another may want an OpenAI model for reasoning. A third may choose a smaller or less expensive model for high-volume tasks. Many large organizations will use several models.
AWS does not need every customer to make the same choice.
It needs customers to make those choices inside AWS.
That distinction is critical.
Models will change. Prices will decline. New architectures will appear. Today’s leader may not dominate every workload three years from now. Bedrock allows Amazon to benefit from model competition while placing AWS at the center of deployment, governance, security, data access and billing.
The models attract attention, but the surrounding infrastructure can create durable customer relationships.
Once a company connects its data, identity controls, monitoring tools, applications and compliance systems to a cloud platform, moving everything elsewhere is not impossible. It is merely unpleasant enough to encourage remarkable loyalty.
Amazon reported that hundreds of thousands of customers were using Bedrock. It added more customers during the six months preceding the second-quarter report than during Bedrock’s first two years, and customers spent more during the quarter than in all previous quarters combined.
That is the type of adoption curve I want to see.
Bedrock is not being preserved behind glass as an innovation exhibit. Customers are using it and spending more money.
Amazon Can Monetize AI Beyond AWS
AWS is the foundation of my thesis, but Amazon’s opportunity extends across the company.
Advertising
Amazon’s advertising business grew 26% in the second quarter. AI can improve ad targeting, campaign creation, product descriptions, recommendations and measurement.
Amazon possesses an advantage that many advertising platforms would happily trade several conference rooms full of executives to obtain: it can observe purchase intent and, in many cases, the completed transaction.
It does not merely know that I watched a video about coffee makers. It may know which coffee maker I bought, when I bought filters and whether I eventually decided that grinding beans manually was too much responsibility before sunrise.
Better AI tools can help merchants create campaigns and help Amazon display more relevant advertising. If those advertisements convert more effectively, merchants may spend more.
Advertising is already a high-margin business. Continued growth can complement AWS and reduce Amazon’s dependence on retail margins.
Retail and logistics
AI can improve inventory placement, demand forecasting, warehouse robotics, delivery routes and customer service.
These applications are less glamorous than a chatbot composing a sonnet, but they may be more valuable.
If AI helps Amazon place the correct product closer to the correct customer, the company can reduce transportation distance and delivery cost. Tiny improvements applied across billions of items can produce meaningful savings.
Amazon reported record Prime delivery speeds during the first half of 2026, with more than 40% additional items delivered the same day or overnight. AI is not solely responsible, but software, automation and forecasting are essential to that network.
Alexa
Alexa spent years answering weather questions, setting timers and occasionally misunderstanding requests with breathtaking confidence.
Generative AI offers Amazon a chance to make Alexa genuinely conversational and useful. Amazon says its upgraded Alexa+ has increased conversation, purchasing, music streaming and smart-home engagement.
The opportunity is real because Alexa already exists across hundreds of millions of endpoints. Amazon does not need to persuade customers to place an entirely new category of device in their homes.
It needs to make the devices already present more valuable.
I remain cautious about the direct financial contribution. Consumer assistants can generate engagement without generating attractive profits. But if Alexa increases purchases, subscriptions or household loyalty, the benefit may appear across Amazon rather than in a single product line.
The $200 Billion Question
Now I arrive at the detail capable of turning optimism into indigestion.
Amazon expects approximately $200 billion in capital expenditures during 2026.
The company spent $128.3 billion in cash capital expenditures during 2025, up from $77.7 billion in 2024. First-quarter 2026 cash capital expenditures reached $43.2 billion, compared with $24.3 billion a year earlier. Amazon’s SEC filings detail the investment increase
This spending has already affected free cash flow.
Amazon’s trailing-12-month operating cash flow increased 33% to $161.4 billion in the second quarter. Normally, that would inspire enthusiastic applause. Yet trailing free cash flow fell to an outflow of $7.6 billion because property and equipment spending increased so sharply, primarily for AI.
Amazon is generating more cash from operations while simultaneously spending even more on infrastructure.
Both statements are true.
This is why I refuse to analyze Amazon using a single cash-flow number without context. Negative free cash flow is not automatically a disaster if the company is investing in assets backed by durable demand and attractive returns. It is also not automatically harmless simply because management calls the spending an investment.
The returns must eventually arrive.
Data centers have long useful lives, but computing equipment becomes obsolete much faster. AI chips, servers and networking systems must produce enough revenue before newer technology reduces their economic value.
The risk is not only that AI demand disappoints. Demand could remain strong while pricing falls, competition rises or hardware advances force additional spending.
Investors are effectively trusting Amazon to allocate one of the largest capital budgets in corporate history.
Amazon has earned some trust. AWS itself was once an enormous, misunderstood investment that became one of the world’s most profitable technology businesses.
Past success, however, is evidence—not immunity.
Competition Will Be Relentless
Amazon is not pursuing this opportunity alone.
Microsoft Azure has deep enterprise relationships and a broad software ecosystem. Google Cloud brings its own AI research, custom chips and model capabilities. Oracle is competing aggressively for large AI workloads. Specialized cloud providers are also building businesses around GPU access.
Then there are the model developers and chip companies, each trying to capture more of the economic stack.
This means AWS must compete on several dimensions at once:
Capacity
Price
Performance
Model selection
Security
Developer tools
Custom silicon
Enterprise relationships
Reliability
AWS has scale and experience, but it does not have a permanent right to win.
Customers may deliberately distribute workloads across multiple providers to improve negotiating leverage and reduce dependence on a single cloud. The AI market can grow rapidly while AWS captures less of that growth than investors expect.
I also expect continuing price pressure. Computing generally becomes cheaper per unit over time. Amazon must lower costs quickly enough to maintain attractive margins as customers demand better economics.
Trainium is central to that defense. If it delivers competitive performance and lowers Amazon’s costs, AWS can reduce prices while protecting margins. If adoption disappoints, Amazon may remain more dependent on expensive third-party hardware.
I Do Not Trust the Headline P/E Ratio
Amazon’s current price-to-earnings ratio appears surprisingly modest for a company producing this level of growth.
There is a reason.
Second-quarter net income reached $62.6 billion, but that included $53.4 billion in pre-tax non-operating income, primarily related to Amazon’s investment in Anthropic.
That gain is economically meaningful, but it is not the same as recurring operating profit from selling cloud services, advertising or merchandise.
When I value Amazon, I do not take the reported trailing earnings figure at face value. I normalize investment gains and focus on operating income, future cash-generating capacity and the earnings power of the underlying businesses.
AWS operating income is real and recurring, although not guaranteed.
An investment valuation gain can disappear or reverse.
This distinction makes Amazon more expensive than the headline P/E ratio suggests, but I still find the long-term valuation reasonable if AWS sustains strong growth and Amazon eventually converts its current capital cycle into substantially higher free cash flow.
My Bull, Base and Bear Cases
I do not believe a single price target can capture every possible outcome, so I prefer scenarios.
Bull case: $370 by the end of 2027
In my optimistic scenario, AWS growth remains above 30% through much of 2027. Trainium adoption expands, Bedrock becomes a central platform for enterprise AI, and operating margins remain in the upper 30% range.
Amazon’s AI infrastructure begins generating revenue faster than new capital spending grows. Free cash flow starts recovering, advertising maintains strong momentum, and investors assign a premium multiple to Amazon’s expanding operating earnings.
Under this scenario, I could justify a share price near $370.
Base case: $315 by the end of 2027
My base case assumes AWS growth moderates from its current 37% rate but remains in the mid-to-upper 20% range. Margins fluctuate as new capacity comes online, and capital spending remains elevated.
Retail margins improve gradually, advertising continues growing faster than the company, and the market gains confidence that Amazon’s AI spending will generate attractive long-term returns.
My 18-month target is approximately $315, representing about 22% upside from the recent price.
Bear case: $205 by the end of 2027
In my bearish scenario, AI infrastructure growth slows sharply, customers delay deployments, pricing pressure intensifies and Amazon’s new capacity takes longer to monetize.
Free cash flow remains weak, depreciation rises, and AWS margins compress. Investors begin treating the capital program as a burden rather than a future earnings engine.
Under those conditions, I could see the stock retreating toward $205.
My Rating: Buy, With Patience
I rate Amazon a Buy for investors with at least a three-to-five-year horizon.
My confidence comes primarily from five factors:
AWS growth has accelerated to 37%.
AWS produces the majority of Amazon’s operating income.
The AI and chips businesses have each surpassed $25 billion annual revenue run rates.
Trainium could improve price-performance and AWS economics.
Amazon can monetize AI across cloud computing, advertising, retail, logistics and consumer devices.
My caution comes from four equally real concerns:
The 2026 capital budget is enormous.
Free cash flow has turned negative under the weight of infrastructure spending.
Competition is intense.
Reported net income is inflated by non-operating investment gains.
I would not chase Amazon after a sudden rally, and I would not build an oversized position based on one quarter. I would accumulate shares gradually, particularly during market pullbacks.
My base-case target is $315 by the end of 2027, with a longer-term path toward higher levels if AWS converts AI demand into sustained revenue, margins and free cash flow.
What I Will Watch Next
I am tracking six indicators.
First, AWS revenue growth. I do not expect 37% forever, but I want growth to remain comfortably above the company-wide rate.
Second, AWS operating margin. Revenue growth is less valuable if depreciation, energy and hardware costs consume the profit.
Third, capital expenditures. I want to know whether spending begins stabilizing after the current construction wave.
Fourth, free cash flow. Amazon can tolerate short-term weakness, but eventually the infrastructure must produce cash.
Fifth, Trainium adoption. Customer commitments are encouraging, but actual utilization and economics matter more.
Sixth, Bedrock usage. I want to see enterprises moving from model experiments into production applications that generate recurring consumption.
If these indicators remain healthy, I will view ordinary stock volatility as noise rather than a broken thesis.
Final Thoughts
Amazon’s AI opportunity is not a side project.
It is becoming one of the central financial questions surrounding the company.
AWS is growing at its fastest rate in years. Its operating income is rising rapidly. Bedrock is gaining adoption. Trainium is attracting major commitments. Amazon’s AI and chip businesses have reached substantial scale.
The company appears to be entering a new infrastructure cycle resembling the early expansion of cloud computing—only larger, faster and dramatically more expensive.
That last part prevents me from becoming careless.
Amazon is spending heavily today for capacity that may not generate significant revenue until 2027 or 2028. Investors must endure weak free cash flow and trust management’s judgment while competitors make similarly aggressive investments.
I am willing to accept that risk, within reason.
AWS has already demonstrated that Amazon can turn massive infrastructure investment into an extraordinarily valuable business. The company now has the customer base, engineering experience, financial capacity and custom technology to compete for a leading position in AI.
I do not know which foundation model will dominate.
I do not know whether today’s leading chip architecture will retain its advantage.
I do not know how quickly autonomous agents will become standard business tools.
What I do know is that nearly every plausible version of the AI future requires computing, storage, networking, security, databases and deployment tools.
AWS sells all of them.
That is why I believe AWS can drive Amazon’s next leg higher. The road will not be smooth, the spending will not be modest, and the market will periodically panic whenever free cash flow looks like it has fallen through an open data-center floor.
But if Amazon converts its current investment into durable AI infrastructure revenue, today’s spending could become the foundation of its next major profit engine.
I am not buying Amazon because artificial intelligence sounds exciting.
I am buying because AWS is beginning to show me the receipts.
Comments
Post a Comment