Cloud Computing: Renting Someone Else's Computer

Why it exists

Before the cloud, launching a digital service required buying, installing, and maintaining physical servers. This involved huge upfront capital costs, long procurement times, and the risk of buying too much or too little capacity. Cloud computing was created to solve this by turning computing infrastructure into a utility, like electricity, that you can consume and pay for as you go.

The mental model

Cloud computing is renting someone else's computers, managed by them, accessible to you over the internet. Instead of owning a car with all its associated costs (purchase, insurance, maintenance, garage), you're hailing a ride-share: you get the transportation you need, when you need it, and only pay for the trip you take. You get access to a fleet of vehicles, from a compact car to a bus, without owning any of them.

How it works

Large providers like AWS, Google Cloud, and Azure build and maintain massive data centers full of servers, storage, and networking gear. They use virtualization to slice up these physical resources into smaller, virtual ones that can be rented by many customers. You use a web dashboard or an API to request resources, and their system automatically provisions and allocates them to you in minutes. Billing is based on your consumption.

When to use it

Use the cloud for nearly any new project, especially when demand is uncertain or variable. It's ideal for startups with limited capital, projects that need to scale up or down quickly (like a viral app or seasonal e-commerce), and for accessing specialized hardware for tasks like machine learning without buying it.

When not to use it

Don't use the cloud for extremely predictable, high-volume workloads that run 24/7 for years, where the cumulative rental cost can exceed the cost of buying your own hardware. It's also not suitable for applications requiring ultra-low latency that can't tolerate network round-trips to a distant data center, or for systems with strict data residency rules.

One canonical example

A startup builds a photo-sharing app. They use a cloud provider to rent a few small virtual servers, a managed database, and an object storage service for the photos. When the app goes viral, they automatically scale up to hundreds of servers to handle the load, then scale back down when traffic subsides, paying only for the extra capacity they used.