We present a proof that Wald's (1947) Sequential Probability Ratio Test for simple hypotheses delivers the smallest possible expected sample size under either hypothesis among all reasonable tests with fixed type I and type II error probabilities.