Antimicrobial resistance (AMR) is one of the greatest global concerns due to the increase in the rate of AMR infections and the lack of development of antimicrobial agents to combat AMR. The development of resistance to carbapenems among common pathogenic bacteria, including Klebsiella pneumoniae and Escherichia coli, is particularly concerning because carbapenems are relied upon for the treatment of Gram-negative infections. Metabolomics offers an approach to identify biomarkers associated with metabolic mechanisms of carbapenem resistance. We analyzed 512 annotated metabolites from a total of 297 bacterial isolates spanning eight organisms. We discovered that the metabolome has greater variation between organisms compared with the resistance phenotype. As a result, we conducted organism-specific statistical analyses to discriminate carbapenem-resistant from carbapenem-sensitive isolates, and supervised learning models resulted in test set mean and standard deviation of area under the receiver operating characteristic curve of 0.822 ± 0.092 and 0.670 ± 0.110, and area under the precision-recall curve of 0.973 ± 0.016 and 0.653 ± 0.121, for Klebsiella pneumoniae and Escherichia coli, respectively. Feature selection using lasso logistic regression identified four metabolite biomarkers of carbapenem resistance in Klebsiella pneumoniae as 6-hydroxyphenazine-1-carboxamide (6-OH-Phz-1-Cam), N-Lactoyl tyrosine (N-Lac tyrosine), flavin adenine dinucleotide (FAD), and amalorin, and seven metabolite biomarkers in Escherichia coli as ibha#30, Cyclo(Leu-Lys), Asukamycin A-II, LPA 16:0, sphinganine, phytosphingosine, and PA 14:0/4:1;O. These unique metabolic signatures provide a vital foundation for exploring the emergence of AMR, warranting follow-up studies to clarify their role in carbapenem resistance and inform improved diagnostics and treatments for AMR.