Switching CostsWide moat

The stickiness of AWS revenue is the heart of its quality as a business, and it comes from the simple, expensive fact that moving is hard — harder with every one of the 200+ proprietary services a workload can entangle itself with¹. Once an enterprise has built its software to run on AWS — wired into its particular databases, its security model, its dozens of specialized services — unwinding all of that to move elsewhere is a project measured in years and fraught with risk, and few companies undertake it without a reason far more compelling than a slightly lower price.

The lock-in has several layers. The applications themselves are written against AWS's specific tools and interfaces. The engineers who built and maintain them are fluent in AWS and would have to relearn a rival's platform. And the data — accumulating into the petabytes over years — is expensive and slow to move, both technically and in the fees a customer pays to extract it. Each layer alone would deter a move; together they make it nearly unthinkable for most.

This is why AWS revenue behaves less like sales that must be won afresh each year and more like an annuity. The customer who signed on years ago keeps spending, and typically spends more over time as its own usage grows, without AWS having to re-win the business. Sticky revenue that expands on its own is the most valuable kind a business can have.

The switching cost also gives AWS pricing latitude. A customer weighing whether the cost of migrating is worth a modest saving almost always concludes it is not, which lets AWS hold prices firmer than a commodity market otherwise would. Lock-in built from a thousand technical entanglements is quieter than a contract, and far stronger.