Value iteration is based on approximating an infinite horizon problem by a finite horizon problem.
max π0 max π1 . . . max πT−1 E [r0 + r1 + · · · + rT−1 + VT (sT )]