Stock Pick: Amazon (AMZN)
Six businesses. One ticker. And the AI boom hasn't even started yet.
Most people think they already know this company.
They’re wrong.
Amazon (NASDAQ: AMZN) stopped being an online bookstore 25 years ago. It stopped being an e-commerce play about 10 years ago. Today, it is something most investors still haven’t fully priced in: a basket of six distinct, high-quality businesses wrapped inside a single ticker. Several of which would be Fortune 500 companies on their own.
And it’s one of the best-positioned enterprises on the planet for the AI boom.
Right now, for the first time in years, it’s trading at a valuation that might actually make sense.
Let’s get into it.
The Business: Not a Company. A Portfolio.
Peter Lynch had a gift for seeing businesses as they actually are, not what they’re labeled as.
If Lynch looked at Amazon today, he wouldn’t call it a retailer. He’d call it what it actually is: six businesses hiding under one ticker symbol. Each one is a mature, cash-generating, and growing business.
Here’s the full breakdown of how Amazon made its $716.9 billion in 2025 revenue:
Business #1: Online Stores
(~37% of revenue | ~$269B)
This is the original engine. Amazon’s first-party marketplace, where Amazon buys products and sells them directly to you. It’s not the fastest-growing segment, but it’s the foundation on which everything else is built. It drives traffic. It trains the algorithms. It anchors Prime membership. Think of it as the loss leader that makes every other business more valuable.
Business #2: Third-Party Seller Services
(~25% of revenue | ~$175B)
Here’s a stat most people don’t know: third-party sellers now account for 62% of all units sold on Amazon. Sure, Amazon sells products. But it also sells access to its platform, collecting commissions, fulfillment fees, and storage fees from over a million independent merchants. This business alone generated roughly $175 billion in 2025.
For context, that’s more than the entire annual revenue of Nike, Starbucks, and McDonald’s combined. Amazon collects a toll on every transaction without owning the inventory.
Business #3: Amazon Web Services (AWS)
(~18% of revenue | ~$129B, ~60% of operating income)
This is where the real money lives.
AWS is the dominant cloud infrastructure platform on Earth, holding roughly 29–30% global market share, exceeding that of Microsoft Azure (~20%) and Google Cloud (~13%). In 2025, AWS generated $128.7 billion in revenue, growing 20% year-over-year, with a 35%+ operating margin.
Think about that for a second. A $129 billion revenue business growing 20% annually at 35% margins. That’s not a division. That’s one of the most valuable standalone businesses in the world; it just happens to be buried inside Amazon’s quarterly report.
AWS represents only 18% of total revenue but generates roughly 60% of all operating income. That math tells you everything.
Business #4: Advertising Services
(~10% of revenue | ~$68.5B)
This is the stealth weapon most investors still underestimate.
Amazon Advertising is now the third-largest digital ad platform on Earth, trailing only Google and Meta. The full-year 2025 advertising revenue hit $68.5 billion and growing nearly 22% year-over-year. In Q4 alone, it generated $21.3 billion.
Here’s the key insight: Amazon’s ads are the most valuable ads in digital media. Google shows you ads when you search. Meta shows you ads while you scroll. Amazon shows you ads when you’re already in the checkout mindset. Purchase intent is the holy grail of advertising. Amazon owns it at scale.
With Prime Video ads now reaching 315 million viewers across 16 countries, this is a media business on top of a commerce business on top of a data business. The margin profile rivals pure software.
Business #5: Subscription Services
(~7% of revenue | ~$49B)
Amazon Prime is one of the greatest subscription businesses ever built. With 260 million members worldwide (185 million in the U.S. alone), Prime bundles free shipping, video streaming, music, photo storage, gaming, and grocery discounts into a single annual fee. The churn rate is remarkably low. Once you’re in the Prime ecosystem, leaving feels expensive even when it isn’t.
Prime Video alone had 240 million viewers in 2025. Amazon spent $22.4 billion on content to keep them there. That’s a content budget that rivals Netflix, and it’s subsidized by the e-commerce flywheel.
Business #6: Physical Stores
(~3% of revenue | ~$21B)
Whole Foods. Amazon Fresh. Amazon Go. This is Amazon’s brick-and-mortar footprint, and it’s quietly becoming a grocery powerhouse. Over 150 million Americans now consider Amazon a primary grocery destination, with same-day perishable delivery expanding rapidly.
Amazon is consolidating around Whole Foods and a new “supercenter” concept to go head-to-head with Walmart in the one retail category that still drives daily physical traffic. Small as a percentage today. Strategically important tomorrow.
The Numbers: Hard to Argue With
Full-year 2025 revenue: $716.9 billion (+12% YoY)
Full-year 2025 net income: $77.7 billion (+31% YoY)
AWS Q4 2025 growth: 24% (the fastest growth in 13 quarters)
AWS annualized run rate: $142 billion
AWS backlog: $244 billion (+40% YoY). This is future locked-in revenue
Operating cash flow: $139.5 billion (+20% YoY)
Advertising revenue Q4 2025: $21.3 billion (+22% YoY)
Prime Video ad-supported audience: 315 million viewers across 16 countries
Trainium AI chips: triple-digit YoY growth, 100,000+ companies using them
These aren’t startup numbers. This is a mature, multi-segment machine that’s accelerating. Not slowing down.
Amazon’s AI Positioning: The Picks-and-Shovels Play of the Decade
Here’s what separates Amazon’s AI story from the rest of the hype cycle:
Amazon doesn’t need to win the AI race. It needs to host it.
Think about the California Gold Rush. The people who got rich weren’t necessarily the miners. They were the ones selling picks, shovels, and denim pants. Amazon, through AWS, is the largest single supplier of picks and shovels to the AI economy, and it’s locking in customers for a decade at a time.
Here’s how each piece of the AI strategy fits together:
Layer 1: The Infrastructure (AWS + Custom Silicon)
Every major AI company (OpenAI, Anthropic, Meta, Adobe, Salesforce, Perplexity, DoorDash, the NBA, the U.S. Air Force) is signing enterprise agreements with AWS. CEO Andy Jassy named them all on the last earnings call. It’s a who’s-who of every sector in the economy.
The deeper story is Amazon’s custom silicon. Instead of buying Nvidia GPUs at a premium, Amazon is building its own:
Trainium2: 1M+ chips deployed, training models at scale, including for Anthropic.
Trainium3: Just launched, 40% better price performance than T2. Nearly all supply committed by mid-2026.
Trainium4: Already in development.
Graviton: 100,000+ customers. Meta just signed on to deploy tens of millions of cores for its agentic AI workloads.
It’s Apple’s M-chip playbook applied to AI infrastructure. Own the silicon, own the margin, own the moat.
Layer 2: The Anthropic Partnership, Amazon’s OpenAI Bet
Amazon has now committed up to $25 billion in total investment into Anthropic, the AI safety company behind the Claude family of models, and one of the most credible frontier AI labs in the world.
In return, Anthropic has committed to spending more than $100 billion on AWS technologies over the next 10 years, training and running its models exclusively on Trainium and Graviton chips. AWS is Anthropic’s primary cloud provider for all mission-critical workloads, including safety research and future model development.
This is a structural lock-in. The more powerful Claude becomes, the more Trainium capacity Anthropic needs. The more Trainium capacity Anthropic needs, the more AWS revenue grows. Anthropic is currently on a $5 billion annualized revenue run rate and growing fast.
And this is just one customer.
Layer 3: Amazon Bedrock, The Enterprise AI Platform
Amazon Bedrock is AWS’s managed AI service that gives enterprise customers access to a curated menu of the world’s best AI models, including Claude (Anthropic), Llama (Meta), Mistral, and Amazon’s own Nova model family, all through a single, secure API.
Bedrock is already a multi-billion dollar annualized run rate business with customer spend growing 60% quarter-over-quarter. The genius of Bedrock is that Amazon doesn’t have to pick the winning AI model. It hosts all of them. When enterprises build on Bedrock, they’re building on AWS infrastructure. Amazon wins regardless of which model they choose.
Layer 4: AI Inside the Retail Business (Rufus + Alexa+)
The AI story extends beyond the cloud. It’s driving measurable revenue in Amazon’s consumer business.
Rufus (AI shopping assistant) was used by 300M+ customers in 2025, generating nearly $12B in incremental annualized sales. Users are 60% more likely to complete a purchase. On Black Friday 2025, AI-assisted sessions ending in a purchase more than doubled vs. the prior month; non-AI sessions grew just 20%. Rufus runs on Anthropic’s Claude and Amazon’s Nova models, meaning Amazon’s AI investment directly drives retail conversion.
Alexa+ users engage twice as often as with the original, and shopping conversations ending in a purchase are up 4x. Four new Echo devices ship natively with Alexa+, putting agentic AI into hundreds of millions of homes.
AI for sellers: 1.3M+ third-party sellers now use Amazon’s gen-AI tools to write listings, optimize pricing, and manage ads. Every efficiency gain increases platform volume and Amazon’s fee revenue.
The AI Summary
Amazon’s AI position is unique because it operates on every level of the stack simultaneously:
At the infrastructure layer: AWS is the dominant cloud, with custom silicon reducing dependency on Nvidia
At the platform layer: Bedrock hosts every major AI model and gives enterprises a single on-ramp to AI
At the application layer: Rufus, Alexa+, and AI seller tools drive measurable revenue today
At the investment layer: Anthropic gives Amazon a strategic stake in one of the top-two frontier AI labs in the world
Most AI companies are competing on one layer. But Amazon is compounding across all four.
The Bull Case: A Flywheel Built for the AI Era
The optimistic case isn’t complicated. It’s just compounding across six businesses, all reinforcing each other.
The capex is a feature, not a bug. $200 billion in planned 2026 capital expenditures sounds alarming until you realize the $244 billion AWS backlog represents signed contracts waiting to be fulfilled. Amazon is building a toll road and the traffic is already booked.
Advertising is still misunderstood. Most investors price Amazon as retail with a cloud division attached. The $68.5 billion advertising business (growing 22%+ at software-like margins) barely registers in their models.
The six-business structure is itself a moat. AWS profits fund retail price leadership. Retail data powers advertising. Advertising funds content. Content locks in Prime. Prime drives more AWS enterprise deals. No competitor can replicate the full system, only pieces of it. And pieces aren’t enough.
The Concerns: What Munger Would Ask
Munger’s inversion rule: “Tell me where I’m going to die, so I don’t go there.”
1. Free cash flow has cratered. With $200 billion in 2026 capex, trailing FCF has dropped to $11.2 billion, down from $38 billion in 2024. The bet is that AWS capacity monetizes as fast as it’s installed. It has historically. But this level of spending leaves no room for error.
2. AWS is losing market share, slowly. At 29–30%, AWS still leads. But it was 33% two years ago. Azure’s OpenAI integration is creating real enterprise stickiness. The margin of dominance is narrowing, even if the absolute lead remains large.
3. Retail margins are structurally thin. North American retail runs at 3–6% margins. Same-day delivery expansion and the new Amazon Leo satellite program (~$1 billion in added 2026 costs) keep pressure on. The profitability story requires AWS and advertising to keep outrunning retail costs. So far, so good, but it requires watching.
4. Valuation demands execution. At ~35x trailing and ~33x forward earnings, there’s no margin of safety in the price. Amazon’s 5-year average P/E was ~48x, so today is relatively cheap by historical standards, but this is still a stock you’re buying on confidence in execution, not cheapness.
The Verdict: This Is the Stock Pickz Buy
Here’s the honest summary:
Amazon is not cheap. It never is.
But what you’re buying at today’s price is meaningfully different from what you were buying three years ago at the same multiple. Back then, Amazon was a company trying to prove its profitability story. Today, that story is proven: $77 billion in net income, $139 billion in operating cash flow, and an AWS segment growing faster than it did a year ago.
The valuation has compressed significantly. Amazon’s forward P/E of ~33x is the lowest it’s been in nearly a decade. Earnings are compounding. The AWS backlog of $244 billion (+40% YoY) means we can see future revenue with unusual clarity.
The flywheel is intact. Retail → Prime → Advertising → AWS. These aren’t separate businesses. They reinforce each other. Every new Prime member makes the advertising platform more valuable. Every ad dollar funds infrastructure. Every infrastructure dollar makes AWS better. This is what Buffett means by a moat: not a wall, but a system that gets harder to compete with over time.
The AI tailwind is real. AWS is the infrastructure layer of the AI economy. Whether the winners of the AI race are OpenAI, Anthropic, Google, or companies we haven’t heard of yet, almost all of them are building on cloud infrastructure. AWS is the dominant provider. Amazon wins as the picks-and-shovels play regardless of which AI model wins.
This is a buy for long-term investors with a 5-10 year horizon.
Disclosure: This newsletter is for educational and informational purposes only and does not constitute financial advice. Always do your own research before making investment decisions.