We explore the evolutionary origins of dosage compensation (DC) in sex chromosomes in the context of metabolic control theory. We consider first the cost of gene loss (hemizygosity) per se in reducing flux, and examine two relationships between flux and fitness (linear and Gaussian) to calculate a fitness cost of hemizygosity. Recognizing that new sex chromosomes are derived from autosomes, we also calculate the cost of unmasking deleterious mutations segregating on the nascent sex chromosomes as loci become hemizygous. The importance of deleterious mutations to the fitness cost of hemizygosity depends on their frequency, and on the relative costs of halving gene dose for wild-type alleles. We then consider the evolution of DC in response to gene loss, and include a cost of overexpression (i.e., DC such that expression exceeds the wild-type homozygote). Even with costs to excess flux, hypomorphic mutations can cause the optimal level of DC to be higher than 2-fold when the absolute cost of hemizygosity is small. Finally, we propose a three-step model of DC evolution: 1) once recombination ceases and the Y begins to deteriorate, genes from longer metabolic pathways should be lost first, as halving these genes does not drastically reduce flux or, thereby, fitness; 2) both the cost of hemizygosity and the presence of hypomorphic mutations will drive an increase in expression, that is, DC; 3) existing DC will now permit loss of genes in short pathways.