Lignocellulosic plant biomass as the substrate for microbial synthesis of 2,3-butanediol is one of the major keys toward cost-effective bio-based production of this chemical at an industrial scale. However, deconstruction of biomass to release the sugars for microbial growth currently requires expensive thermochemical and enzymatic pretreatments. In this study, the thermo-cellulolytic bacterium Caldicellulosiruptor bescii was successfully engineered to produce 2,3-butanediol from cellulose, xylan, and directly from unpretreated switchgrass. Genome-scale metabolic modeling of C. bescii was applied to adjust carbon and redox fluxes to maximize productivity of 2,3-butanediol, thereby revealing bottlenecks that require genetic modifications.