Amazon

In the spring of 1994, a thirty-year-old vice president at the hedge fund D.E. Shaw came across a number he could not stop thinking about: internet usage was growing at 2,300 percent a year. Jeff Bezos made a list of twenty things a person might sell online, ranked them, and landed on the least glamorous item on the page. Books. Millions of titles existed, no physical store could shelve them all, and they shipped without breaking. He drove west, incorporated a company in a Bellevue garage, and rigged a bell to ring every time an order came in. Within weeks the bell had to be switched off. It never stopped ringing.

Thirty-two years later, the bookstore posts two hundred billion dollars of revenue in a single quarter. It runs the computing backbone for a large slice of the internet, operates one of the planet's biggest advertising businesses, streams the NFL, employs roughly 1.55 million people, and has put more than 300 satellites into orbit. Understanding Amazon means holding one uncomfortable idea in your head: it is not really a retailer anymore. It is a machine for compounding, and retail was simply the first thing fed into it.

The store that became a mall that became a toll road

Start with what most people see: the everything store. Amazon.com sells books, batteries, groceries, sofas, and nearly anything else that fits in a cardboard box, to hundreds of millions of customers in dozens of countries. The problem it solved in 1995 was selection - no bookstore could stock every book. The problem it solves in 2026 is friction. You think of a thing, and the thing appears at your door, often the next day, sometimes the same one.

But the store's most consequential decision came in 2000, when Amazon invited outside merchants - competitors, effectively - to sell on its own pages. Analysts thought it was self-sabotage. It turned out to be the whole plot. Today roughly 60 percent of units sold on Amazon come from third-party sellers, millions of independent businesses that pay Amazon commissions, fulfillment fees, and, increasingly, advertising fees to be seen. The store became a mall. The mall became a toll road. Amazon collects whether or not its own products win.

Wrapped around all of it is Prime, launched in 2005 as a $79-a-year bet that Wall Street openly hated. Free two-day shipping for a flat fee looked like a money incinerator. Instead it became arguably the most effective loyalty program in retail history - more than 200 million members who get shipping, movies, music, and grocery discounts in one bundle, and who, once subscribed, rarely shop anywhere else first.

"Put the customer first. Invent. And be patient." Jeff Bezos, on the three ideas behind Amazon's success

The side project that pays for everything

Here is the twist that separates Amazon from every retailer it is compared to: the profits mostly do not come from retail. In the early 2000s, Amazon's engineers were rebuilding the same infrastructure - storage, servers, databases - for every new project. They standardized the pieces, and in 2006 someone asked the question that changed the company: what if we rented this out?

Amazon Web Services effectively invented the modern cloud computing industry. Startups no longer needed to buy servers; they needed a credit card. Netflix, Airbnb, government agencies, and a majority of large enterprises came to run on AWS. Two decades on, it remains the market leader ahead of Microsoft Azure and Google Cloud, and in the second quarter of 2026 it grew 37 percent - its fastest pace in eighteen quarters - to an annualized run rate of about $169 billion, with operating margins near 40 percent.

That acceleration has one word behind it: AI. AWS now sells the picks and shovels of the artificial intelligence boom - data centers, custom Trainium chips for training models, Graviton processors for general computing, and Bedrock, a marketplace of AI models. Amazon has invested more than $8 billion in Anthropic, the maker of the Claude models, with commitments of up to $25 billion more, and has agreed to supply up to five gigawatts of computing capacity to train and run Claude. Five gigawatts is the output of several nuclear reactors, dedicated to one customer. The company that once optimized cardboard boxes now optimizes electrons.

The napkin diagram that explains everything

If you want the entire strategy on one page, it already exists - Bezos sketched it on a napkin in Amazon's early days. Lower prices attract customers. More customers attract sellers. More sellers mean more selection and scale. Scale lowers the cost structure. Lower costs allow lower prices. The loop feeds itself, and every new Amazon business is judged by one test: does it spin the wheel faster?

The flywheel explains businesses that otherwise look random. Kindle (2007) removed friction from reading. Echo and Alexa (2014) removed friction from ordering. Whole Foods ($13.7 billion, 2017) put the flywheel in the produce aisle. Prime Video keeps members renewing. Fulfillment by Amazon rents the logistics machine to sellers, the way AWS rents the computing machine to developers. Even the advertising business - now $19.8 billion in a single quarter, quietly one of the largest ad operations on Earth - is a toll on the traffic the flywheel generates.

And then there is Amazon Leo, formerly Project Kuiper: a constellation of low-Earth-orbit broadband satellites, more than 300 of them in orbit by mid-2026, the third-largest constellation in space, with consumer service rolling out across the US, UK, France, Germany, and Canada. Why would a retailer build satellites? Because every human newly connected to the internet is, eventually, a customer. Starlink sees a rival; the flywheel just sees more spokes.

"We still won't have enough capacity to meet all the demand we have in 2026." Andy Jassy, CEO, after raising capital spending to roughly $220 billion

Six pages, two pizzas, and Day 1

Amazon's expertise is usually described as logistics or computing, but its strangest advantage is procedural. PowerPoint is banned in meetings. Anyone proposing something significant must write a six-page narrative memo, in full sentences, which the room reads silently for the first twenty minutes. The theory: bullet points hide sloppy thinking, and prose exposes it. New products begin with a mock press release written before a line of code exists - if the announcement is not exciting, the product is not built. Teams are kept small enough to be fed by two pizzas.

Above it all sits the "Day 1" doctrine from Bezos's 1997 shareholder letter, reattached to every annual report for decades: Day 2 is stasis, followed by irrelevance, followed by death. The culture licenses spectacular failure as tuition. The Fire Phone flopped and lost hundreds of millions; its team and lessons rolled directly into Echo and Alexa, which created an entire product category. Bezos's own standard was blunt - if the size of your failures is not growing, you are not inventing hard enough.

The doctrine cuts in harsher directions too. Between late 2025 and mid-2026, Amazon eliminated roughly 30,000 corporate roles, with Jassy citing AI-driven efficiency and too many layers of management. A company that preaches Day 1 to shareholders applies it, without much sentiment, to its own org chart. Its warehouse workforce, the largest in American private industry after Walmart, has long been a flashpoint for unions and regulators, and its scale keeps antitrust lawyers on both sides of the Atlantic permanently employed. The machine compounds; the friction is human.

Where it sits, and who it fights

Because Amazon is several companies wearing one trench coat, it has no single competitor - it has one per limb. Walmart, Alibaba, Shopify, and the discount insurgent Temu fight it in retail. Microsoft and Google chase AWS in the cloud, and both posted strong numbers while AWS was in its slower 2025 stretch, before Amazon's AI buildout swung momentum back. Google and Meta dwarf most ad rivals, but Amazon's ads convert at the moment of purchase - the bottom of the funnel, where money actually changes hands. Netflix and Disney contest the living room. Starlink owns the sky Amazon Leo is entering. UPS and FedEx now watch their former customer deliver billions of its own packages.

No competitor faces Amazon everywhere, which is precisely the point. Squeeze the retail margin and AWS pays the bills. Pressure the cloud and advertising picks up the slack. The conglomerate structure that looks unfocused from the outside functions as a portfolio of mutually reinforcing tolls - and as one of history's largest internal funding mechanisms, currently channeling about $220 billion a year into data centers, chips, and rockets.

For everyone else, Amazon is less a company than a set of utilities. Shoppers use it as the default price check for the physical world. A student with an idea can rent world-class infrastructure by the hour. A candlemaker in Ohio can reach 200 million Prime members without hiring a single driver. An author can publish to a global bookstore overnight through Kindle Direct Publishing. An AI lab can train frontier models on gigawatts of capacity it never had to build. You do not have to admire Amazon to depend on it, and that dependency, more than any quarterly number, is the measure of what got built.

The logo tells you the plan was always this. The yellow smile under the wordmark is an arrow, running from A to Z. Everything, in other words. In 1994 that was a bookstore's ambition. In 2026 it reads like an inventory list still being written - groceries to gigawatts, doorbells to orbit, A to Z and then some distance past Z, delivered, if the flywheel holds, by tomorrow morning.