On-demand
Paying for compute as you use it without long-term commitment — expensive but flexible.
Think of it like
Renting a car for one day instead of buying it.
Example
Spin up an inference server for an hour, use it, shut it down; pay only for that hour.
How it actually works
On-demand is the default in cloud computing. You pay more per unit than reserved capacity, but you avoid the risk of overcommitting. For bursty workloads (spiky traffic), on-demand is better. For steady workloads, reserved capacity saves money.
For product teams
On-demand is flexible but expensive; switch to reserved capacity if usage is predictable.
For engineers
On-demand is great for dev/test; production workloads should use reserved or spot.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome