Sure, but the altitude gain is almost "free" - you need to accelerate by something like 6 km/s laterally (number pulled out of somewhere, it's 2.3 in kerbal space program but the planets are smaller there) and 1 km/s vertically (depending on ascent profile, 0 for an instantaneous ascent and infinite for a slow hover upwards, but it ends up something like that), so by Pythagoras it's sqrt-37 = really not a lot more than the 6 kms/s you'd have to spend if you could start at 100km altitude.
Intuitively we assume rockets go "upwards" partly because that's the part we see, but actually the gravity turn occurs very early and most of the acceleration is closer to horizontal. E.g. the less-crazy of the space shuttle abort plans if something went wrong early in flight was to land at Shannon.
The Tsiolkovsky equation suggests that's going to be a big saving:
http://en.wikipedia.org/wiki/Tsiolkovsky_rocket_equation
And yes, my only qualification is Kerbal as well ;-)