The ability of multiple servers to work in conjunction and provide redundancy is important in the cloud, especially for large data centers. One value of such a system is its failover capabilities -- if any server is incapacitated, another one can take over for it. The servers can also provide readily available backups that can scale to immense sizes.
Within a farm, there are several services that run on one or more servers. These services provide basic functionality for SharePoint and regulate which services should run on which servers, in an effort to manage the impact on overall farm architecture and performance.
There are occasions when multiple SharePoint farms make sense. An enterprise might have a quality assurance (QA) farm, for example, that is separate from a production farm. Geographically dispersed enterprises might have a farm in Europe, one in the Asia-Pacific region and another in North America. Finally, when different parts of an organization have differing polices regarding customization and support of SharePoint, it makes sense to have separate farms. It is then possible for the enterprise to replicate or share services as needed.