split sklearn's IncrementalPCA (ipca.transform) input, based on features instead of samples

Question

I have a huge input for IncrementalPCA on number of features: (22, 258186260) I wonder if its okay to split sklearn's IncrementalPCA (ipca.transform) input based on features instead of samples. Would the result still be correct.

               ipca = IncrementalPCA(n_components=2)
               for batch in np.split( input, 10,axis=1):
                     
                     output=ipca.transform(batch)