Amazon
The flywheel and its costs
Description
In July 1995, a company operating out of a rented garage in Bellevue, Washington shipped its first order: a book on computational neuroscience, sold through a website with a search box and not much else. The company was called Amazon, after the river, chosen partly because its founder wanted something that suggested vastness and partly because it started with an A and would land near the top of alphabetical lists. Jeff Bezos had left a comfortable job at a New York hedge fund to sell books online, a decision he later framed as the choice he would least regret at eighty. For a while, the plan looked modest. Books were easy to ship, easy to catalogue, and there were more of them than any physical store could hold.
Three decades later, Amazon is one of the largest companies on earth, employing well over a million people, running a cloud-computing business that quietly powers a large slice of the internet, and delivering packages fast enough that same-day has stopped feeling like a miracle. The bookseller became the store for everything, then the infrastructure underneath other stores, then a logistics network, a film studio, a grocery chain, an advertising giant. What held all of it together was not a product but a shape — a self-reinforcing loop that Bezos and his colleagues came to call the flywheel, a machine designed to spin faster the more it was fed.
That machine is genuinely elegant, and it is worth understanding on its own terms rather than as a villain or a marvel. It explains why Amazon kept lowering prices when rivals raised them, why it went years without meaningful profit and investors kept cheering, why it expanded into businesses that seemed to have nothing to do with each other. But a machine built to grow without a natural stopping point pushes against the people and markets it touches. That pressure is where the story gets complicated.
The question we’re asking : How did a loop drawn to sell books more cheaply turn into one of the most powerful — and most contested — companies in the world?What we’ll see : How the flywheel was assembled, what it swallowed as it spun, and the weight it puts on the people and markets in its path.
Table of contents
01Chapter 1 — The letter that promised no profits
In 1997, Amazon went public and Bezos wrote his first letter to shareholders — a document he would attach to every annual report that followed. It made a promise most companies avoid: Amazon would not chase quarterly profits. It would spend everything it could on growth, market share, and long-term position, and it asked investors who wanted short-term returns to look elsewhere. This was not modesty. It was a strategy dressed as candour, and it set the terms for everything after.
The logic ran against instinct. A normal retailer wants margin — the gap between what it pays and what it charges. Amazon treated margin as something to give away. Every efficiency it found, every dollar it saved on warehousing or shipping, went back into lower prices rather than into the bank. The bet was that low prices would pull in customers, customers would pull in more sales, more sales would let Amazon negotiate harder with suppliers and spread its fixed costs wider, and those savings would fund still lower prices. Round and round.
02Chapter 2 — The flywheel, drawn on a napkin
The image that came to define Amazon's strategy has a modest origin. In the early 2000s, in a conversation involving Bezos and the management writer Jim Collins, whose work on great companies Amazon's leaders admired, the team sketched a loop on a napkin. Lower prices bring more customers. More customers attract more third-party sellers onto the platform. More sellers widen selection and improve the customer experience. A better experience brings still more customers. And greater scale lowers the cost structure, which funds lower prices again. No single point was the engine. The whole ring was.
What made the flywheel more than a slogan was that Amazon actually restructured itself around it. The launch of Marketplace in 2000, letting outside merchants sell alongside Amazon's own inventory, seemed counterintuitive — inviting competitors onto your own shelves. But through the flywheel it made sense: every outside seller added selection Amazon did not have to buy, and every sale, whether Amazon's or a third party's, fed the same loop. Today more than half of the goods sold on Amazon come from these outside sellers, each paying fees for the privilege of reaching Amazon's customers.
03Chapter 3 — The everything store becomes the everything company
The most consequential expansion was one customers rarely see. In 2006, Amazon began selling the computing power it had built for itself — servers, storage, databases — to other companies, under the name Amazon Web Services. The insight was that running warehouses of computers well was hard, and Amazon had gotten good at it out of necessity. Why not rent that capacity out? AWS grew into the backbone of a huge share of the modern internet; startups, banks, streaming services and government agencies now run on it. Crucially, it is highly profitable, and for years its earnings quietly funded the thin margins of the retail business.
With cash flowing, the loop pointed everywhere. Amazon built its own logistics arm — planes, trucks, sorting centres — to stop depending on carriers and to squeeze delivery times toward same-day. It bought Whole Foods in 2017 for about thirteen billion dollars, planting a flag in physical grocery. It made devices, the Kindle and the Echo, less to earn money on hardware than to keep customers inside its orbit. It became one of the largest advertising businesses in the world, selling the top of its own search results back to the sellers competing on it.
04Chapter 4 — What a machine built to grow keeps growing into
A flywheel has no natural stopping point. That is its genius and its problem. Every gain gets reinvested to make the next gain easier, which means the pressure to be faster, cheaper, and larger never eases — it compounds. When that pressure meets a human workforce, it lands as intensity. Amazon's warehouses run on tracked productivity: pick rates, time between scans, targets that ratchet as the system learns what is achievable. Reporting over the past decade, along with regulatory findings in several countries, has documented injury rates above industry averages and turnover so high that in some regions Amazon has cycled through much of the available local labour pool. The efficiency the customer feels as convenience is, further up the line, felt as strain.
The same logic reshapes the markets Amazon operates in. Third-party sellers depend on a platform that is also their competitor: Amazon sees what sells, sets the fees, controls the search ranking, and has at times introduced its own versions of products doing well on the site. Sellers describe it, not unfairly, as renting a store from a landlord who studies your receipts. Regulators in the United States and the European Union have opened cases probing exactly this — whether a company that runs the marketplace can fairly compete inside it. The Federal Trade Commission's 2023 lawsuit put the question directly, alleging Amazon uses its dominance to keep prices higher and sellers captive.
05Conclusion
The book on computational neuroscience that left the Bellevue garage in 1995 was, in a literal sense, just an early customer order. In another sense it was the first turn of a wheel that has not slowed since. Everything that followed — Marketplace, Prime, AWS, the planes and the grocery stores and the advertising — was the same loop, fed and re-fed, doing what it was designed to do: get bigger, cheaper, faster, and use each gain to buy the next. The elegance of the design is not in dispute. Neither, increasingly, is its weight.
Related books
