Computational prediction of protein function is frequently error-prone and incomplete. In Mycobacterium tuberculosis (Mtb), ~25% of all genes have no predicted function and are annotated as hypothetical proteins, severely limiting our understanding of Mtb pathogenicity. Here, we utilize a high throughput, quantitative, activity-based protein profiling (ABPP) platform to probe, annotate, and validate ATP-binding proteins in Mtb. We experimentally validate prior in silico predictions of >250 proteins and identify 72 hypothetical proteins as novel ATP binders. ATP interacts with proteins with diverse and unrelated sequences, providing a new and expanded view of adenosine nucleotide binding in Mtb. Several hypothetical ATP binders are essential or taxonomically limited, suggesting specialized functions in mycobacterial physiology and pathogenicity.