Note that this assume [tex]p\neq1[/tex]; if that were the case, we'd end up with a logarithm as the antiderivative, and it's easy to show that in that case the integral diverges.
We have two cases remaining. If [tex]p<1[/tex], then the numerator is of degree larger than 0, and as [tex]x\to\infty[/tex] we have [tex]x^{1-p}\to\infty[/tex].
If [tex]p>1[/tex], then the degree of the numerator will be negative, which would mean the numerator approaches 0 as [tex]x\to\infty[/tex], leaving us with just [tex]\dfrac1{1-p}[/tex]. So the integral only converges for [tex]p>1[/tex].