I played it through just to see if I could, and it did look a bit rough in the middle there. At one point, my command tank had lost all of it's defensive units, and was holding off a few enemy cyborgs by itself and limping back home while it took near-forever for reinforcments to arrive. Fortunately, it did make it, as I would not have been able to afford to replace it.
At the most abstract level, the concept is ok, but I would think that you would want something of a timer for the player to "win" if they have been able to hold out for so long; not sure if or how you can implement that in this sort of a setting. I'm not sure if the stock AI will do a good job of launching mass attacks - usually what I see is scattered groups of small # of units cruising through to pull off an attack. Dydo-ai might do better. You can lock down the starting settings (bases, power, AI strength) via a .ini file (as was done for the 3 included challenges) but that's probably not called for until it's a bit more playable.
Map: I think the map is probably too large in the vertical dimension; it makes for very long supply lines and I don't think the stock AI does well with that. The map is featureless in the middle, which is a bit boring. If the oil in base was buildable, it probably wouldn't be that hard to hold out for near-indefinite time periods. I would suggest less in-base oil, with perhaps some outside base oil near the AI players defensive line. A more pronounced hill might be a bit nice as well, maybe with cliff edges?
That way the human player has a natural advantage of the hill for defense, but the AI player starts with a superior position and
I think it will take more playtesting to really get this one down.