It is one of the types of hierarchical clustering.
In theory, it can be done by initially grouping all the observations into one cluster, and then successively splitting these clusters. This is known as divisive hierarchical clustering.
This approach is also known as the Top-Down Approach.