Utility computing
Utility computing is a model for buying computing resources, such as processing power and storage, as a metered service. Customers pay for measured usage rather than buying and maintaining enough hardware to cover their maximum needs.
How it works
A provider supplies capacity and sets billing units: instance-hours, gigabyte-months of storage or gigabytes of data transferred. Your bill follows those units, much like an electricity bill.
Metered usage is not always productive work. A running virtual machine can incur charges even when your app has no users. Stopping it might end compute charges while charges for its stored data continue.
For a startup, the benefit is avoiding a large hardware purchase before demand is known. The trade-off is a variable bill: more customers, inefficient code or unexpected traffic can increase costs.
How a small app produces a $27 bill
A small app uses these resources during one month. Rates are illustrative, not a provider quote. The example excludes taxes, free tiers and minimum charges.
| Resource | Monthly usage | Illustrative rate | Cost |
|---|---|---|---|
| Virtual machines | 2 machines × 200 hours = 400 hours | $0.05/hour | $20 |
| Storage | 100 GB-months | $0.02/GB-month | $2 |
| Outbound data transfer | 50 GB | $0.10/GB | $5 |
| Total | $27 |
The arithmetic is $20 + $2 + $5 = $27. Doubling compute hours would add $20, not double the entire bill.
Before choosing a provider, check what starts and stops each meter. Compare the whole expected bill, not just the advertised compute rate.
Questions people ask
What is the difference between cloud computing and utility computing?
Cloud computing describes how resources are delivered: on-demand access over a network to a shared pool of resources. Utility computing describes how customers pay. They often overlap, but they are not synonyms. NIST's cloud definition includes metering; it does not require a particular billing model.
