K-means vs. K-means++

In this Lesson we are introduced to the advantages of using K-means++ with respect to K-means clustering.
There are few things that puzzle me.
Nowhere is said what the drawbacks of using K-means++ instead of the k-means clustering method are.
Can there be performance issues in using the ++ method when the dataset is very large?
It also appears to me, but I might be wrong, that there are only specific edge cases, with a very particular arrangement of the data points, where the ++ method actually leads to a model improvement.