The ability of plant-associated microbes to produce gibberellin A (GA) phytohormones was first described for the fungal rice pathogen Gibberella fujikuroi in the 1930s. Recently the capacity to produce GAs was shown for several bacteria, including symbiotic alpha-proteobacteria (α-rhizobia) and gamma-proteobacteria phytopathogens. All necessary enzymes for GA production are encoded by a conserved operon, which appears to have undergone horizontal transfer between and within these two phylogenetic classes of bacteria. Here the operon was shown to be present and functional in a third class, the beta-proteobacteria, where it is found in several symbionts (β-rhizobia). Conservation of function was examined by biochemical characterization of the enzymes encoded by the operon from Paraburkholderia mimosarum LMG 23256T. Despite the in-frame gene fusion between the short-chain alcohol dehydrogenase/reductase and ferredoxin, the encoded enzymes exhibited the expected activity. Intriguingly, together these can only produce GA9, the immediate precursor to the bioactive GA4, as the cytochrome P450 (CYP115) that catalyzes the final hydroxylation reaction is missing, similar to most α-rhizobia. However, phylogenetic analysis indicates that the operon from β-rhizobia is more closely related to examples from gamma-proteobacteria, which almost invariably have CYP115 and, hence, can produce bioactive GA4. This indicates not only that β-rhizobia acquired the operon by horizontal gene transfer from gamma-proteobacteria, rather than α-rhizobia, but also that they independently lost CYP115 in parallel to the α-rhizobia, further hinting at the possibility of detrimental effects for the production of bioactive GA4 by these symbionts.