It is theoretically possible to engineer Saccharomyces cerevisiae strains in which isobutanol is the predominant catabolic product and high-yielding isobutanol-producing strains are already reported by industry. Conversely, isobutanol yields of engineered S. cerevisiae strains reported in the scientific literature typically remain far below 10 % of the theoretical maximum. This study explores possible reasons for these suboptimal yields by a mass-balancing approach. A cytosolically located, cofactor-balanced isobutanol pathway, consisting of a mosaic of bacterial enzymes whose in vivo functionality was confirmed by complementation of null mutations in branched-chain amino acid metabolism, was expressed in S. cerevisiae. Product formation by the engineered strain was analysed in shake flasks and bioreactors. In aerobic cultures, the pathway intermediate isobutyraldehyde was oxidized to isobutyrate rather than reduced to isobutanol. Moreover, significant concentrations of the pathway intermediates 2,3-dihydroxyisovalerate and α-ketoisovalerate, as well as diacetyl and acetoin, accumulated extracellularly. While the engineered strain could not grow anaerobically, micro-aerobic cultivation resulted in isobutanol formation at a yield of 0.018±0.003 mol/mol glucose. Simultaneously, 2,3-butanediol was produced at a yield of 0.649±0.067 mol/mol glucose. These results identify massive accumulation of pathway intermediates, as well as overflow metabolites derived from acetolactate, as an important, previously underestimated contributor to the suboptimal yields of ‘academic’ isobutanol strains. The observed patterns of by-product formation is consistent with the notion that in vivo activity of the iron-sulfur-cluster-requiring enzyme dihydroxyacid dehydratase is a key bottleneck in the present and previously described ‘academic’ isobutanol-producing yeast strains.