With the increasing prevalence of multi-drug resistant bacteria, last-resort antibiotics like ceftazidime-avibactam and cefiderocol have become crucial. Due to the lack of rapid diagnostic tools and their absence in most automated antibiotic susceptibility tests (AST), empirical initiation of therapy is common for critically ill patients. We propose a rapid phenotypic AST using the Resistell Phenotech, measuring bacterial nanomotions – or vibrations - within two hours. The growth-independent AST was developed and tested on clinical E. coli and K. pneumoniae isolates, recording nanomotions for ceftazidime-avibactam and cefiderocol. Machine learning algorithms achieved predictive models with 93% accuracy for spiked positive blood cultures, providing results within two hours. This rapid and accurate nanomotion AST challenges the conventional delayed gold-standard tests in hospitals, potentially transforming the assessment of antibiotic susceptibility and informing treatment decisions faster, thereby curbing the spread of antimicrobial resistance for critical drugs.