For several decades, the conventional wisdom has been that expected firm growth rates are independent of firm size, a property known as Gibrat's Law. However, recent empirical work has found a negative relation between firm growth and firm size. This paper provides a theoretical explanation for this negative relation in a model of new firm growth where capacity and technology choices involve some degree of sunkness. Additional empirical implications of the theory of sunk costs are also developed. These relate to profit rates, Tobin's Q, exit rates, degree of sunkness of capacity costs, as well as size and growth rates.