Non-obstructive azoospermia (NOA) is the most severe form of male infertility and typically incurable with current medicine. Due to the biological complexity of sperm production, defining the genetic basis of NOA has proven challenging, and to date, the most advanced classification of NOA subforms is based on simple description of testis histology. In this study, we exome-sequenced over 1,000 clinically diagnosed NOA cases and identified a plausible recessive Mendelian cause in 20\%. Population-based testing against fertile controls identified 27 genes as significantly associated with azoospermia. The disrupted genes are primarily on the autosomes, enriched for undescribed human “knockouts”, and, for the most part, have yet to be linked to a Mendelian trait. Integration with single-cell RNA sequencing of adult testes shows that, rather than affecting a single cell type or pathway, azoospermia genes can be grouped into molecular subforms with highly synchronized expression patterns, and analogs of these subforms exist in mice. This analysis framework identifies groups of genes with known roles in spermatogenesis but also reveals unrecognized subforms, such as a set of genes expressed specifically in mitotic divisions of type B spermatogonia. Our findings highlight NOA as an understudied Mendelian disorder and provide a conceptual structure for organizing the complex genetics of male infertility, which may serve as a basis for disease classification more advanced than histology.