What is Cloud Computing?

Cloud computing means renting computers, storage, and services over the internet instead of buying and running your own hardware. Someone else (a cloud provider like AWS) owns the data centers, the servers, the cooling, and the network. You ask for what you need, you use it, and you pay only for what you used — by the second or by the gigabyte. This page explains the idea in plain terms, when the cloud is the right choice, and the one cost trap that surprises most teams.

The old way vs the cloud way

In the traditional (“on-premises” or “on-prem”) model, you buy physical servers, install them in a room or rented data center, and look after everything: power, cooling, network cables, operating system patches, and hardware replacement. You pay for all of it up front, whether or not you use it.

With cloud computing you do the opposite. You request a virtual server, it appears in a minute or two, and the meter starts. Turn it off and the meter stops. This is the heart of on-demand computing: capacity is there when you ask, and gone when you don’t.

The key shift is from CapEx (capital expenditure — a big up-front purchase you own) to OpEx (operational expenditure — a running cost you rent). You trade owning assets for paying as you go.

Aspect	On-premises (own it)	Cloud (rent it)
Up-front cost	High (buy hardware)	Almost zero
Time to launch	Weeks to months	Minutes
Scaling up	Order, ship, install	One API call
Scaling down	You still own the box	Stop paying instantly
Who patches hardware	You	The provider
Billing	CapEx (own the asset)	OpEx (pay per use)

What you actually rent

Cloud providers sell building blocks. The three you will meet first on AWS:

Compute — virtual servers. On AWS this is EC2 (Elastic Compute Cloud), which gives you a virtual machine you control.
Storage — durable space for files and data. On AWS this is S3 (Simple Storage Service) for files and EBS (Elastic Block Store) for disks attached to servers.
Networking — private networks, IP addresses, and load balancers that tie it all together.

This style — renting raw infrastructure you configure yourself — is called IaaS (Infrastructure as a Service). It is the most flexible model and the closest to running your own servers. There are higher-level models too, covered in Cloud service models.

When the cloud wins

The cloud shines when your needs are uneven, urgent, or worldwide.

Variable load. A shopping site is quiet at 3 a.m. and slammed on Black Friday. In the cloud you add servers for the rush and remove them after. You only pay for the peak while it lasts. On-prem, you would have to buy enough hardware for the peak and let it sit idle the rest of the year.
Fast launch. A new project needs a server today, not after a 6-week purchase order. The cloud delivers in minutes.
Global reach. AWS has data centers around the world (see Regions and Availability Zones). You can run your app close to users in Tokyo, Frankfurt, and Virginia without owning anything in those places.
Experiments. Spin something up, try it, delete it. If it fails you paid only for the hours you used.

Tip: “Scalability” is the cloud’s headline feature. Elastic scaling means capacity automatically grows and shrinks with demand, so you are not paying for idle machines at night.

When on-premises still makes sense

The cloud is not always the answer.

Steady, predictable load. A system that runs flat-out 24/7 every day of the year can be cheaper to own. Rented capacity carries a margin; if you never turn it off, you pay that margin forever.
Strict data residency. Some laws or contracts require data to physically stay in a specific building or country under your control. On-prem (or a tightly controlled private setup) can make compliance simpler.
Specialized or legacy hardware. Equipment that cannot run on standard cloud servers may have to stay in your own racks.

For a deeper side-by-side, see Cloud vs on-premises.

The “pay as you go” gotcha

“Pay as you go” sounds like it is always cheaper. It is not. Per-hour cloud pricing includes the provider’s profit and the convenience of instant scaling. If you run a server at high utilization around the clock for years, owning the hardware can cost less in total.

Warning: At steady, high utilization, on-demand cloud pricing can cost more than buying hardware. The cloud rewards variable load and punishes always-on waste. For long-running steady workloads, use Reserved Instances or Savings Plans — commit to 1 or 3 years of usage in exchange for up to ~72% off the on-demand rate.

A quick, concrete example: a general-purpose m7i.large server runs roughly $0.10 per hour on-demand in the US East region. Run it 24/7 for a month and that is about $74. The same server on a 3-year Savings Plan can drop to around $30/month. Multiply that gap across a fleet and the difference is real money.

See it in action

You don’t need a real server to feel the on-demand model. Listing what AWS can offer in a Region takes one command. First, see how to set up the CLI in Install the AWS CLI.

Using the AWS Management Console:

Sign in to the AWS Management Console.
In the top search bar, type EC2 and open the service.
Click Launch instance — notice you can pick a server size and have it running in about a minute, with no purchase order.

Using the AWS CLI (v2), check what instance types are available before you commit to anything:

aws ec2 describe-instance-type-offerings \
  --location-type availability-zone \
  --filters "Name=instance-type,Values=t3.micro" \
  --region us-east-1

Output:

{
    "InstanceTypeOfferings": [
        {
            "InstanceType": "t3.micro",
            "LocationType": "availability-zone",
            "Location": "us-east-1a"
        },
        {
            "InstanceType": "t3.micro",
            "LocationType": "availability-zone",
            "Location": "us-east-1b"
        }
    ]
}

Nothing was bought, nothing was billed — you simply asked the provider what is on the menu. That request-and-response, billed only when you actually run something, is cloud computing in one line.

Best Practices

Match the model to the workload: cloud for variable or bursty load, owned/committed capacity for steady 24/7 load.
Turn off what you are not using — idle resources still bill in the cloud.
For predictable long-running workloads, buy Reserved Instances or Savings Plans instead of paying on-demand rates.
Set up billing alerts on day one so a forgotten server cannot quietly run up a bill.
Right-size before you scale: a smaller instance that is busy beats a large one that is mostly idle.
Keep data-residency rules in mind when choosing where to run — pick a Region that satisfies your legal requirements (see Choosing a Region).