Many exploits, particularly buffer overflows, need to be carefully tailored to the hardware the victim is running on. They work to insert executable binary code into places it would not ordinarily be found. It doesn't do an attacker much good to feed x86 code to an m68k processor. At worst, they can crash the program, leading to a denial of service, but they can't break in.
Tailoring the 'payload' of an exploit is fairly sophisticated work that can't be performed by just any random script kiddie. It requires fairly intimate knowledge of the program, operating system and hardware. It's extremely rare that a cracker will devote the time to generating a Linux/Sparc or NetBSD/Alpha payload - they almost always stick to Linux/Intel.