What type of instances allows users to pay only for the compute time they consume?

On-demand instances provide a flexible and cost-effective option for users who want to pay only for the compute time they consume. This type of instance allows users to launch and terminate instances according to their needs without any long-term commitment. Users are charged on an hourly or per-second basis depending on the instance type, which is ideal for workloads with unpredictable usage patterns or those that cannot be interrupted.

This model is particularly beneficial for applications that require flexibility and scalability, as users can quickly adapt to changing demands without being tied to a subscription or reservation. Unlike reserved instances, which involve an upfront commitment for a lower rate over time, on-demand instances allow for immediate use and payment strictly based on usage.

While spot instances offer reduced rates by allowing customers to bid on unused EC2 capacity, they come with the risk that the instance can be terminated if the spot price exceeds the bid amount. Dedicated instances provide physical servers dedicated to a single customer's use, which does not align with the concept of paying solely for consumed compute time since costs are incurred regardless of usage. Reserved instances help save costs over a longer commitment but require an upfront payment and are not billed based solely on usage.