LABORATORY INVESTIGATION OF BACTERIAL INFECTIONS GENERALLY TAKES TWO DAYS: one to grow the bacteria and another to identify them and to test their susceptibility. Meanwhile the patient is treated empirically, based on likely pathogens and local resistance rates. Many patients are over-treated to prevent under-treatment of a few, compromising antibiotic stewardship. Molecular diagnostics have potential to improve this situation by accelerating precise diagnoses and the early refinement of antibiotic therapy. They include: (i) the use of 'biomarkers' to swiftly distinguish patients with bacterial infection, and (ii) molecular bacteriology to identify pathogens and their resistance genes in clinical specimens, without culture. Biomarker interest centres on procalcitonin, which has given good results particularly for pneumonias, though broader biomarker arrays may prove superior in the future. PCRs already are widely used to diagnose a few infections (e.g. tuberculosis) whilst multiplexes are becoming available for bacteraemia, pneumonia and gastrointestinal infection. These detect likely pathogens, but are not comprehensive, particularly for resistance genes; there is also the challenge of linking pathogens and resistance genes when multiple organisms are present in a sample. Next-generation sequencing offers more comprehensive profiling, but obstacles include sensitivity when the bacterial load is low, as in bacteraemia, and the imperfect correlation of genotype and phenotype. In short, rapid molecular bacteriology presents great potential to improve patient treatments and antibiotic stewardship but faces many technical challenges; moreover it runs counter to the current nostrum of defining resistance in pharmacodynamic terms, rather than by the presence of a mechanism, and the policy of centralising bacteriology services.