Yep. Shader loading on ATI cards take a lot longer than on Nvidia cards. This is a very real problem which I think ATI needs to address. In Just Cause 2 you'll probably see maybe 2 seconds load time from startup screen to frontend menu if you're using an Nvidia card, whereas the same takes maybe 10 seconds if you're on an ATI card. The difference is all about shader loading time.
HLSL to D3D asm can be precompiled, but the optimization step is done by the driver at runtime. A driver could opt to cache already optimized shaders though, but I don't think anyone does.
Can ATI actually fix this problem via drivers? This is something I've seen mentioned for few generations now.