Forecasting Back Through a Difference

Monthly revenue $y(t)$ at a store drifts upward, so it is non-stationary. An analyst takes first differences,

$\Delta y(t) = y(t) - y(t-1)$

and finds that the differenced series is well described by an AR(1) model:

$\Delta y(t) = 2 + 0.5 \, \Delta y(t-1) + \varepsilon(t), \qquad E[\varepsilon(t)] = 0$

The two most recent observed months are

$y(T-1) = 140, \qquad y(T) = 150$

with revenue measured in thousands of dollars.

What is the point forecast $\hat{y}(T+2)$ — the revenue level two months after the last observation?