Cereulide and isocereulides A-G are biosynthesized as emetic toxins by Bacillus cereus via a non-ribosomal peptide synthetase (NRPS) called Ces. Although a thiotemplate mechanisms involving cyclo-trimerization of ready-made D-O-Leu-D-Ala-L-O-Val-L-Val via a thioesterase (TE) domain is proposed for cereulide biosynthesis, the exact mechanism is far from being understood. UPLC-TOF MS analysis of B. cereus strains in combination with (13)C-labeling experiments now revealed tetra-, octa-, and dodecapeptides of a different sequence, namely (L-O-Val-L-Val-D-O-Leu-D-Ala)1-3, as intermediates of cereulide biosynthesis. Surprisingly, also di-, hexa-, and decadepsipeptides were identified which, together with the structures of the previously reported isocereulides E, F, and G, do not correlate to the currently proposed mechanism for cereulide biosynthesis and violate the canonical NRPS biosynthetic logic. UPLC-TOF MS metabolite analysis and bioinformatic gene cluster analysis highlighted dipeptides rather than single amino or hydroxy acids as the basic modules in tetradepsipeptide assembly and proposed the CesA C-terminal C* domain and the CesB C-terminal TE domain to function as a cooperative esterification and depsipeptide elongation center repeatedly recruiting the action of the C* domain to oligomerize tetradepsipeptides prior to the release of cereulide from the TE domain by macrocyclization.