k-means clustering, is an algorithm that assigns n observations to exactly one of k clusters defined by centroids, where k is chosen before the algorithm starts [1].
It means that even we run k-means with k = 400/20 = 20, it does not guarentee that each cluster has equal number of elements.
If you want to ensure that each cluster has equal number of elements, you may run a post-processing step, wherein you reassign datapoints between clusters to balance the cluster sizes. One approach is to iteratively reassign the data points from the largest clusters to the smallest clusters until all clusters have the desired number of elements.
