I cant get it.
Best used for non-critical, temporary workloads that are fault-tolerant. For example, you may need some instances to do batch jobs, process a few photos or videos, or testing, etc.
Read more: https://go.aws/3K8TSkm
To elaborate on Ricardo, lets say you are a creator and need to encode a video to multiple formats. Being the good producer that you are, you know you need to upload multiple formats to your website, and you allow one week buffer time after editing to do this work and stage a video before its made public. You could save money by setting this encoding work on a batch job. Then at some point this week (likely late at night or something) when demand for compute power and corresponding prices drop, your staged spot instances will launch and this work can be done for you automatically at a significant savings for you.