Gustafson's Law posits that the potential speedup of a parallel computing system is determined by the proportion of the problem that can be parallelized, rather than the fixed size of the problem. It emphasizes scalability, suggesting that as more processors are added, larger problems can be solved in the same amount of time, which contrasts with Amdahl's Law that focuses on fixed workloads.