Existence of a loop in a Turing machine?


Consider a Turing Machine which (1) reads all its input and (2) accepts inputs arbitrarily large. Can we conclude that there must be a loop in the finite-state control as its inputs get larger?