When Amazon launched the AWS EC2 cloud computing service back in 2006, per-hour billing was a big deal, but that scheme also meant that you'd pay for a full hour even if you only used an instance for a few minutes. Over the last few years, AWS's competitors moved to more flexible billing models (mostly per-minute billing) and now, starting October 2, AWS is one-upping many of them by moving to per-second billing for its Linux-based EC2 instances.
This new per-second billing model will apply to on-demand, reserved and spot instances, as well as provisioned storage for EBS volumes. Amazon EMR and AWS Batch are also moving to this per-second model. it's worth noting, though, that there is a one-minute minimum charge per instance and that this doesn't apply to machines that run Windows or some of the Linux distributions that have their own separate hourly charges.
In a world where "serverless" platforms like AWS Lambda are quickly taking off -- and where you also only pay per second -- the old per-hour model surely didn't make much sense to a lot of developers. But there's also the simple fact that EC2 has gotten a bit of a reputation for being expensive, especially as Google, Microsoft and others moved to per-minute billing.