Amazon

Amazon is not one company but three interlocking businesses sharing a single culture: the retail store, the third-party Marketplace, and Amazon Web Services (AWS). Each was built to serve Amazon first, then opened to outsiders, and the stories below trace how that sequence, powered by a deliberately low-margin core and a profit engine hidden in higher-margin layers, turned each business into a structural advantage rivals cannot copy piece by piece.

Business Model: financing a low-margin retail empire with a high-margin cloud

The problem. Retail is a brutally low-margin business, and Amazon chose to make it lower still, running the retail core at razor-thin or negative margins to keep prices down and competitors out. That raised a hard question: how do you fund relentless expansion (warehouses, delivery, new categories) when the core business barely profits by design?

The approach. Amazon concentrated profit in higher-margin layers and used them to subsidize the low-margin core. Chief among them is AWS, the internal infrastructure Amazon built for itself and then sold to everyone else, alongside third-party seller services and advertising. The cloud cash cow pays for retail's expansion.

How it solved it. In Q4 2020, AWS generated roughly $12.7 billion in revenue and about $3.56 billion in operating income (up 37% year over year), representing only around 10% of Amazon's sales but roughly 52% of its total operating income. That single high-margin segment funds retail growth that would otherwise be unaffordable, letting Amazon keep prices low without starving investment.

Moats: turning tech-company profits into hard-to-copy concrete and trucks

The problem. Any single advantage can be copied. A rival can match a price, clone a feature, or undercut a category. Amazon needed advantages that get harder to replicate the bigger they grow, not easier.

The approach. Amazon poured its high-margin profits into physical infrastructure: fulfillment centers, sortation hubs, an air cargo network, and last-mile delivery. The capital intensity is the point. Building a nationwide logistics network is expensive precisely because it is hard to copy, what Ben Thompson called "real moats built with real dollars."

How it solved it. A competitor cannot beat Amazon by winning one node; it must replicate the entire network simultaneously, at similar scale and cost, which requires years and billions. Third-party sellers, who account for roughly 58% to 60% of units sold, plug into that same fulfillment network (via Fulfilled by Amazon), so every dollar Amazon spends on logistics deepens the moat for the retail store, the Marketplace, and the sellers all at once.

Disruptive Innovation: writing the press release before writing the code

The problem. As Amazon grew, the natural drift of a large company is toward internal, engineering-driven, feature-first thinking, building what is technically interesting rather than what customers actually want. Bezos framed the antidote as staying "Day 1," treating the company as perpetually early to resist the complacency of scale.

The approach. Amazon institutionalized invention through "Working Backwards." Before a team builds anything, it writes a roughly six-page PR/FAQ: a mock press release announcing the finished product from a future date, plus a frequently-asked-questions section. Teams start by defining the customer experience and work backwards to what to build, killing weak ideas on paper before a line of code is written.

How it solved it. The process forced customer-obsessed clarity across most of Amazon's major launches since 2004, including Kindle, Prime, and AWS itself, as documented by former Amazon VPs Colin Bryar and Bill Carr in Working Backwards. By shifting the starting point from "what can we build" to "what will the press release say," it repeatedly converted vague internal capability into products defined by customer benefit.

Network Effect: opening the store to rival sellers to widen selection

The problem. No single retailer can stock everything, and Amazon's own first-party inventory carried both capital cost and the risk of unsold goods. Thin selection sends customers elsewhere, but infinite first-party inventory is financially impossible.

The approach. Amazon opened its store to third-party sellers, letting outside merchants (even direct competitors) list on the same product pages as Amazon itself. More sellers bring broader selection, which draws more customers, whose traffic attracts still more sellers: a two-sided loop that is a core node of Bezos's flywheel, first sketched on a napkin around 2001.

How it solved it. Third-party sellers grew from under 3% of units sold in 1999 to roughly 58% to 60% by 2020, and Amazon monetizes them through fulfillment fees, seller services, and a fast-growing ads business without carrying their inventory risk. The result is near-limitless selection funded largely by other people's capital, with each new seller making the store more valuable to buyers and each new buyer making it more valuable to sellers.