Updated 26 October 2026 · Published by Zoorik
A burstable instance is a cloud virtual machine (VM) with a baseline CPU level. It earns CPU credits while it runs below that baseline, then spends them to burst above it.
Kubernetes’ Burstable QoS class is a different thing: a pod class with no credits.
Burstable instances in AWS and Azure
- A burstable instance in AWS is a T instance, such as T3 or T4g.
- A burstable VM in Azure is a B-series VM, such as Bsv2.
- CPU credits in AWS measure virtual CPU (vCPU) time. One credit is one vCPU at 100% for one minute.
- T3 unlimited mode is the default. A T3 on a Dedicated Host can’t use it.
Out of credits, an Azure B-series VM or an AWS instance in standard mode falls back to its baseline. In AWS unlimited mode, the instance keeps bursting.
A t3.large past its baseline
A t3.large has 2 vCPUs and a baseline of 30% per vCPU. AWS’s example sets it to unlimited and runs it at a 24-hour average of 40% CPU. The hourly price covers the first 30%, and the extra 10% is charged at a flat rate per vCPU-hour.
Why it matters for your cloud cost
- A steady load above the baseline runs slower, or costs extra in unlimited mode.
- At AWS’s example prices, a t3.large costs less than a same-size fixed-CPU m5.large, up to a 42.5% daily average CPU. Above that, it costs more.
- Check the baseline when you pick a node size.
The free assessment works out what your clusters and disks could save.
Read next: Kubernetes node size best practices: large or small nodes
Sources: Amazon Web Services, “Burstable performance instances” (opens in a new tab) (Amazon EC2 User Guide); Amazon Web Services, “Key concepts for burstable performance instances” (opens in a new tab); Amazon Web Services, “Unlimited mode concepts for burstable instances” (opens in a new tab); Microsoft Learn, “B family VM size series” (opens in a new tab).