Amdahl's law
Speedup from parallelisation is bounded by the serial portion — if 10 % of the work must run serially, the max speedup is 10×, no matter how many cores you throw at it.
Formulated by Gene Amdahl in 1967. Mathematically: speedup(N) = 1 / ((1-P) + P/N), where P is the parallelisable fraction and N is the number of processors. Explains why "throw more cores at it" has its limits and why engineers should optimise the serial portion first. Complement: Little's law, Gustafson's law (more optimistic if the problem itself grows with the processor count).