If you are new to cloud computing you probably already understand the concept of provisioning servers and only paying for the capacity you are using at an hourly rate. This “on demand” concept for computing requires that a certain amount of excess capacity is available at all times to ensure that customers are able to provision the number of services they need for their applications. With this in mind, cloud services like AWS and Azure must always have a certain percentage of excess capacity for all of their data centers.
AWS pioneered the concept of a “server spot market” to allow customers to purchase this excess capacity at a steep discount (up to 90). The Spot Market provides a “market price” for spot instances based on the server region, availability zone, and type/size. This market price for instances is in constant flux and can go anywhere from 10% of the on-demand price all the up to 10x the on-demand price. If you bid too low, your instance will be terminated prematurely. If you bid too high you can potentially pay much more than the list on demand price. You can see why provisioning servers in the spot market is not a viable option for most cloud customers.
You can learn more about the AWS EC2 Spot market from the AWS website
November 2017 Update:
During Amazon's latest re:Invent Amazon EC2 simplified the Spot pricing by moving to a model which delivers low, predictable prices that adjust gradually, based on long-term trends in supply and demand. In addition, bidding became optional.