Cloud Infrastructure: Components, Costs and How to Choose

Cloud infrastructure is the rented hardware and software that your website or app runs on: servers, storage, networks, databases and the security controls around them, all owned and operated by a provider and paid for as you use them. If you are a founder choosing where to host a product, or a freelance web developer setting up a client site, the hard part is not the definition. It is knowing which layer you actually need, what it will cost once real traffic arrives, and which security jobs stay with you.

This guide covers what cloud infrastructure is made of, the main service models (IaaS, PaaS, serverless and managed platforms), how it differs from cloud computing and cloud architecture, the main providers, real starting prices, a decision tree for a first stack, and the mistakes that cause surprise bills.

Every price and free-tier limit below was checked on the providers' official pricing pages on November 9, 2026, in US dollars before tax. Cloud pricing changes often and varies by region, so confirm on the provider's page before you commit.

What is cloud infrastructure?

Cloud infrastructure is the collection of physical and virtual resources that a cloud provider runs in its data centers and offers over the internet: compute, storage, networking, databases and identity and access controls. You never touch the physical machines. You create resources through a web console, a command-line tool or an API, and they are ready in seconds or minutes.

The US National Institute of Standards and Technology (NIST) defines cloud computing in its publication SP 800-145 as a model for on-demand network access to a shared pool of configurable computing resources, such as networks, servers, storage, applications and services, that can be rapidly provisioned and released with minimal management effort. Cloud infrastructure is the "shared pool" part of that definition: the resources themselves.

In practice, three things make it different from buying your own server:

• On demand. You create a server or a database when you need it and delete it when you don't.
• Pay as you go. You are billed by the hour, second, request or gigabyte instead of paying up front for hardware.
• Elastic. You can scale up for a launch and back down afterwards, often automatically.

The core components of cloud infrastructure

Almost every cloud product, from a giant AWS account to a simple Vercel project, is built from the same six building blocks.

Compute

Compute is the processing power that runs your code. It comes in several forms:

• Virtual machines (VMs). A virtual server with its own operating system, such as Amazon EC2, Google Compute Engine, Azure Virtual Machines or a DigitalOcean Droplet.
• Containers. Your app packaged with its dependencies, run on a service such as Google Cloud Run or Kubernetes.
• Serverless functions. Small pieces of code that run only when a request arrives, such as AWS Lambda, Vercel Functions or Cloudflare Workers.

Storage

Storage keeps your files and data. The three types you will meet are object storage (files and images in "buckets", such as Amazon S3, Google Cloud Storage or Cloudflare R2), block storage (a virtual disk attached to a VM) and file storage (a shared network drive).

Networking

Networking connects everything and lets users reach it. It includes virtual private networks, firewalls, load balancers, DNS and public IP addresses. When someone types your URL, DNS turns the domain into an IP address and the provider's network routes the request to your app.

Databases

Most apps need a database. In the cloud you can run one yourself on a VM, or use a managed database where the provider handles installs, backups and patching. Examples include Amazon RDS, Google Cloud SQL, Azure SQL Database and Supabase (managed Postgres).

Content delivery network (CDN)

A CDN stores copies of your static files (images, CSS, JavaScript and often whole pages) on servers around the world, so visitors download them from a location near them. Vercel, Cloudflare and Amazon CloudFront are common choices. For most websites, a CDN does more for speed than a bigger server.

Identity and access management (IAM)

IAM controls who and what can do things in your account: which team member can deploy, which app can read a storage bucket, and which API key can call a service. Weak IAM is behind many cloud security problems, so it is a core component, not an extra.

Cloud infrastructure examples

Here is what cloud infrastructure looks like for a few common projects:

• A marketing site or blog. Static pages on a CDN, a form handled by a serverless function, and DNS. No server to manage.
• A SaaS app. A framework such as Next.js running on a managed platform, a managed Postgres database, object storage for uploads, and an authentication service.
• A WordPress site. One virtual server with a web server, PHP and MySQL, plus backups and a CDN in front.
• An API for a mobile app. Containers or serverless functions behind an API gateway, a managed database and a queue for background jobs.

IaaS vs PaaS vs serverless vs managed platforms

Cloud infrastructure services are usually grouped by how much the provider manages for you. NIST names three service models: Infrastructure as a Service (IaaS), Platform as a Service (PaaS) and Software as a Service (SaaS). Serverless and modern "deploy from Git" platforms sit between IaaS and PaaS, or on top of PaaS.

A simple rule: the further down the IaaS line you go, the more control you get and the more work you take on. DigitalOcean's own pricing FAQ describes its Droplets as IaaS: you get virtualized compute, storage and networking, and you are responsible for the operating system, applications and data.

Cloud infrastructure vs cloud computing vs cloud architecture

These three terms get mixed up constantly. Here is the difference:

• Cloud computing is the overall model: renting computing resources over the internet on demand. It includes SaaS apps you simply log in to.
• Cloud infrastructure is the set of resources underneath: the servers, storage, networks, databases and IAM that cloud services run on.
• Cloud architecture is the design: how you combine those resources for a specific app, such as which region, how many servers, where the database sits, how traffic flows and how failures are handled. It is usually drawn as a cloud infrastructure diagram.

Compared with on-premise infrastructure, where you buy and run servers in your own office or data center, cloud infrastructure swaps up-front hardware costs and maintenance for an ongoing bill and less control over the physical layer.

Main cloud infrastructure providers

There are two broad groups of providers. Which one fits depends on how much of the stack you want to own.

The big three hyperscalers

Amazon Web Services (AWS), Microsoft Azure and Google Cloud offer hundreds of services in regions across the world, including India (AWS lists an Asia Pacific (Mumbai) Region, for example). They can run anything, but their consoles, IAM models and bills take time to learn. Their free offers for new customers on November 9, 2026 were:

• AWS: $100 in credits when you create a Free Tier account, plus up to $100 more for exploring key services, for up to $200 over 6 months on the Free plan. AWS says the Free plan account closes after 6 months or when credits run out, and 30+ services are always free within monthly limits.
• Google Cloud: a $300 Welcome credit to use over 90 days, plus an always-free tier that includes one e2-micro VM per month in us-west1, us-central1 or us-east1, 30 GB-months of standard persistent disk, and 2 million Cloud Run requests per month.
• Microsoft Azure: a $200 credit to use within 30 days, free monthly amounts of 20+ services for 12 months (including 750 hours each of B1s, B2pts v2 and B2ats v2 burstable VMs) and 65+ always-free services.

Developer platforms and simpler clouds

Vercel, Cloudflare, DigitalOcean, Supabase and AWS Lightsail package infrastructure into fewer, simpler products. You give up some flexibility, but you can launch in an afternoon and the bill is easier to predict. Many of them run on top of, or alongside, the hyperscalers.

What cloud infrastructure costs: real starting prices

Cloud bills are made of a few cost drivers: compute time, storage, data transfer out (egress), requests, managed database size and paid seats. The table below shows official starting prices for common building blocks.

What a small stack costs: an illustrative demo

To show how these numbers add up, here are three example stacks priced from the list prices above. They are illustrative demos, not real bills, and they exclude tax, domains, email and usage beyond the included amounts.

• Managed platform stack: Vercel Pro ($20) plus Supabase Pro ($25) comes to $45 a month for one developer seat and one database project, before overages.
• Simple VPS stack: an AWS Lightsail 1 GB Linux instance ($7), a Lightsail standard managed database ($15) and a Lightsail CDN distribution ($2.50) comes to $24.50 a month.
• Edge stack: Cloudflare Workers Paid ($5) with R2 storage inside the 10 GB free amount comes to $5 a month at low usage.

For a broader budget that includes design and development, read how much a website costs, scope by scope, try the free Website Cost Calculator, and see website maintenance cost for what you keep paying after launch.

The cost drivers people miss

• Egress (data transfer out). Sending data to users usually costs more than storing it. AWS includes 100 GB a month of free data transfer out across services; Lightsail object storage overage in US East is $0.09 per GB. Cloudflare R2 and Workers do not charge for egress.
• Public IPv4 addresses. AWS charges $0.005 per hour for each public IPv4 address, in use or idle. That is $3.60 for a 30-day month per address.
• Idle resources. A test server, database or load balancer you forgot about keeps billing. A Lightsail load balancer, for instance, is $18 a month whether or not it gets traffic.
• Region differences. Prices and allowances vary by region. Lightsail plans in Asia Pacific (Mumbai) include half the data transfer allowance of the same plan in US regions.
• Seats and add-ons. Platform plans often charge per developer, and features like password protection, extra backups or log retention cost more.

How to choose: a decision tree for your first cloud stack

Use this decision tree to pick a starting point. Answer the questions in order and stop at the first match.

1. Is it a mostly static website (marketing site, portfolio, blog, docs)?
   • Yes: use a managed web platform with a CDN, such as Vercel or Cloudflare Pages. Check the plan terms: Vercel Hobby is for personal, non-commercial use, so a business site needs Pro.
   • No: go to question 2.
2. Is it a web app built with a framework like Next.js, with users and a database?
   • Yes: use a managed platform for the app plus a managed database such as Supabase. See Next.js vs React if you haven't picked a framework yet.
   • No: go to question 3.
3. Is it an API, webhook handler or background job with spiky or low traffic?
   • Yes: use serverless, such as Cloudflare Workers, Google Cloud Run or AWS Lambda, so you pay little when it is idle.
   • No: go to question 4.
4. Does it need a long-running server or software you must install yourself (WordPress, a game server, a custom worker, a self-hosted tool)?
   • Yes: use a simple VPS such as a DigitalOcean Droplet or AWS Lightsail, and plan who patches the operating system.
   • No: go to question 5.
5. Do you need specialist services (GPUs, data warehouses, strict compliance, private networking with an enterprise customer)?
   • Yes: use a hyperscaler (AWS, Azure or Google Cloud), ideally with someone who has worked with its IAM and billing before.
   • No: start with the simplest managed option above and move later only when you hit a real limit.

Two tie-breakers help when you're unsure. First, choose the provider your team already knows, because unfamiliar consoles cause more downtime than slow servers. Second, pick a region close to most of your users; for an Indian audience that might be a Mumbai or Bangalore region, both of which appear on the AWS and DigitalOcean pages above.

Cloud infrastructure security and the shared responsibility model

Cloud infrastructure security is split between you and the provider. AWS calls this the shared responsibility model: AWS handles security "of" the cloud, meaning the hardware, software, networking and facilities that run its services, while the customer handles security "in" the cloud.

How much falls on you depends on the service model. AWS gives two clear examples. With Amazon EC2, which is IaaS, the customer manages the guest operating system (including updates and security patches), any installed software and the security group firewall rules. With more abstracted services such as Amazon S3 and DynamoDB, AWS operates the infrastructure, operating system and platform, and the customer is responsible for its data, encryption options, asset classification and IAM permissions.

The same idea applies to every provider. A basic security checklist for a small project:

• Turn on multi-factor authentication for every account, especially the owner or root login.
• Give each person and each app only the permissions it needs, and use separate keys for development and production.
• Never put secret keys in front-end code or a public Git repository. Store them in the platform's environment variables or secret manager.
• Keep storage buckets private unless a file must be public.
• Patch the operating system on any VM you run, or choose a managed service so the provider does it.
• Turn on automatic backups for databases and test a restore at least once.
• Set budget alerts so an attack or a bug that loops forever shows up as an email, not a surprise invoice.

Cloud infrastructure management basics

Cloud infrastructure management is the ongoing work of keeping resources running, secure and affordable. For a small team it comes down to a few habits:

• Infrastructure as code. Define servers, databases and DNS in configuration files (for example with Terraform or the provider's own tools) so you can recreate them and review changes in Git.
• Monitoring and logs. Watch uptime, errors and response times, and keep logs long enough to debug an incident.
• Cost controls. Use budgets and alerts. Vercel gives new teams a default on-demand budget of $200 with notifications, and you can set it to pause projects at a hard limit. Supabase spend caps are on by default on Pro. Google Cloud offers billing budgets and AWS has AWS Budgets.
• Regular clean-up. Once a month, delete test resources, old snapshots and unused IP addresses.
• Documentation. Write down which accounts exist, who owns them and how to deploy. A client or co-founder should be able to find this without you.

AI coding assistants can draft infrastructure-as-code files and explain error messages, but review their changes the way you would review any code. Our guide to AI tools for web development covers which tools suit this kind of work.

Common cloud infrastructure mistakes

• Starting on a hyperscaler for a simple site. A brochure site doesn't need a VPC, load balancer and container cluster. Start simple and move when you have a reason.
• Treating free tiers as production. Free plans pause, cap or expire. Supabase Free projects pause after a week of inactivity, Google's $300 credit ends after 90 days and Azure's $200 credit after 30 days.
• Ignoring plan terms. Vercel Hobby is for personal, non-commercial use. Check the terms before launching a business on a free plan.
• No budget alerts. Without them, the first sign of a problem is the invoice.
• Using the root or owner account day to day. Create separate users with limited permissions.
• Forgetting egress and IP charges. Storage looks cheap until a popular video or large download is served from a provider that bills per gigabyte.
• Putting the database in a different region from the app. Every query then crosses a long network path, which slows pages and can add transfer charges.
• No exit plan. Keep your code in Git, your database exportable and your DNS at a registrar you control, so you can switch providers if prices or terms change.

FAQ

Next steps

Start with the decision tree, pick the simplest option that fits, and set a budget alert on day one. For the building blocks behind it, read what an API is, what an IP address is and what a URL is. If you want help planning a stack or building the site itself, get in touch.

Official sources

Prices and limits checked November 9, 2026: NIST SP 800-145 · AWS Free Tier · AWS Shared Responsibility Model · Amazon EC2 on-demand pricing · Amazon VPC pricing · AWS Lightsail pricing · Google Cloud free features · Azure account · Vercel pricing · Cloudflare Workers pricing · DigitalOcean Droplet pricing · Supabase pricing.