Your desired result can't possibly be what you really want, it's not even the right length.
I'm guessing you made the plot using gscatter and wondered why the "obvious" clusters in your data aren't what kmeans returned. But look at the scaling of your two data columns: the range of the first one is four orders of magnitude larger than the second. By having such poorly scaled variables, you are in effect ignoring the second one completely, and you can see that the clusters that kmeans finds are in fact the same as if you had used just the first column of A.
Run A through something like perhaps zscore to standardize it, and you'll get what I suspect you're looking for.
