Okay, [itex]A_n= 1.0075A_{n-1}- 4000[/itex]
A standard method for something like [itex]A_n= rA_{n-1}[/itex] is to try something like [itex]A_n= n^x[/itex] for some number. If that were true, then [itex]A_{n-1}= (n^x)^r= n^{rx}[/itex] and [itex]A_n= 1.0075A_{n-1}[/itex] becomes [itex]n^r= 1.0075 (n^{r-1})[/itex]. Dividing both sides by [itex]n^r[/itex], [itex]1= 1.0075r^{-1}[/itex] so r= 1.0075. In fact, if we were to try [itex]A_n= C(1.0075)^n[/itex], for C any constant, we would have [itex]A_n= C(1.0075)^n= 1.0075A_{n-1}= 1.0075C(1.0075)^{n-1}= C(1.0075)^n[/itex] is true for all n because the "C"s cancel. [itex]C(1.0075)^n[/itex] is the general solution to the equation [itex]A_n= 1.0075A_n[/itex].
That's ignoring the "-4000" part but since that number is a constant, what if we try [itex]A_n= A[/itex], a constant? Now [itex]A_n= 1.0075A_{n-1}- 4000[/itex] becomes A= 1.0075A- 4000 or -.0075A= 4000.
Now the "theory" part: If [itex]A_n[/itex] is the general solution to the "homogeneous" equation and A is a single solution to the entire equation, then [itex]A_n+ A[/itex] is the general solution to the entire equation.
You should now be able to write out the general solution, use the fact that [itex]A_1= 400000[/itex] to find C and then determine when [itex]A_n= 0[/itex]. (You may find that it is never 0. Just find when it is less than 1.)