Amazon Web Service has announced Amazon Elastic Compute Cloud per second billing

Amazon Web Service introduced Amazon Elastic Compute Cloud usage of the Linux usage based instances that are launched in On-Demand, Spot, and reserved form will be billed on one-second accretion with a minimum of 60 seconds. 

Amazon Elastic Compute Cloud Elastic GPUs and Amazon Elastic BookStore volumes will also transition from per hour billing to per second billing with a minimum 60 second. As with the Elastic Compute Cloud instances the list prices will continue to be displayed as it is currently in GB month of capacity, per Million I/O requests for Magnetic IOPS and IOPS-month provisioned for io1 IOPS. 

Customers may have to utilize cases that need a large number of instances operating irregular periods of time such as the data processing, batch processing, dev/test, analytics, gaming applications and image rendering. To increase the flexibility and enable customers to pay more focus and attention to the application instead of maximising the usage to the hour with the Elastic Compute Cloud per second billing, the customers can now stop, start and terminate the EC2 instances. This will provide comfort in the fact that their EC2 resources will be not charged for the seconds and minutes which are remaining in the hours the instances were operating.  

This feature is available in all the Amazon Web Service Regions and is effective immediately for Linux instances that launched newly or are already running. Per-second billing is not available currently to instances that are operating on the Linux distribution or Microsoft Windows that has a separate hourly charge.