Eigenvectors represent fixed directions, not fixed magnitudes. From Wikipedia:

> More precisely, an eigenvector v of a linear transformation T is scaled by a constant factor lambda when the linear transformation is applied to it: Tv = lambda v .

In other words, repeated multiplication of an eigenvector by a matrix can still create exponential growth.