Cloud computing explained for non-engineers

The original cloud provider and the most feature-rich. Dominant in enterprises across every industry. If a job description lists cloud experience and does not specify a provider, AWS is the most likely requirement.

Strong in enterprises already using Microsoft products — Office 365, Active Directory, Teams. If you are joining a large enterprise that runs on Windows and Microsoft infrastructure, it is probably on Azure.

Strong in data and AI — BigQuery for analytics, Vertex AI for machine learning. If you are joining a data-heavy company or an AI-first startup, GCP is common. Google's networking infrastructure is also considered best-in-class for global latency.

Virtual machines — the 'servers' your app runs on. Instead of a physical machine in a data center, you get an allocated slice of a real server. You pay per hour of use. Need more capacity during a product launch? Spin up more instances. Traffic slows down? Scale back and stop paying.

Object storage for files — images, videos, backups, data exports. Very cheap at scale. When users upload a profile photo or your app exports a CSV report, those files go to object storage. AWS S3 (Simple Storage Service) is the most widely used; Azure Blob Storage and Google Cloud Storage are the equivalents.

Managed databases — PostgreSQL, MySQL, and others. The cloud provider handles backups, patches, scaling, and failover. You focus on your data, not on database administration. 'Managed' is the key word: instead of a DBA keeping the database running, the cloud does it automatically.

Content delivery network — serves static files (images, CSS, JavaScript) from servers physically close to your users for fast load times. If your users are in Europe and your servers are in Virginia, a CDN caches your assets in Frankfurt. The page loads from 20 miles away instead of 4,000.

Code that runs on demand, without managing servers. No server running 24/7 — you pay only when code executes. A common use case: an image resize function that triggers when a user uploads a photo. It runs for 200 milliseconds, then stops. You are billed for those 200 milliseconds, not for a full server hour.

Packaged apps that run consistently anywhere. A container bundles the app and all its dependencies into a single unit. This eliminates 'it works on my machine' — the container runs identically in development, staging, and production. Kubernetes (K8s) is the orchestration layer that manages many containers at scale.

Region

Geographic data center location (us-east-1, eu-west-2). Data sovereignty laws affect which regions companies can use — EU companies often cannot store user data outside Europe due to GDPR. Latency also matters: users in Asia-Pacific should be served from an APAC region.

Availability Zone (AZ)

Multiple physically separate data centers within a region. Distributing your app across AZs means that if one data center has a power outage, your product stays live on the others. High availability architectures always span multiple AZs.

SLA (Service Level Agreement)

The uptime guarantee from the cloud provider. AWS S3 SLA is 99.9% — that is roughly 8.7 hours of potential downtime per year. AWS EC2 SLA is 99.99% — about 52 minutes per year. These numbers matter when you are designing a product with uptime requirements in your contract with customers.

IAM (Identity and Access Management)

Who has permission to do what in the cloud account. IAM controls which engineers can deploy to production, which services can read from the database, and which automated processes can access secrets. Misconfigured IAM is one of the most common causes of cloud security incidents.