A new approach is presented for cell lysate identification which uses SERS-active silver nanoparticles and a droplet-based microfluidic chip. Eighty-nanoliter droplets are generated by injecting silver nanoparticles, KCl as aggregation agent, and cell lysate containing cell constituents, such as nucleic acids, carbohydrates, metabolites, and proteins into a continuous flow of mineral oil. This platform enables accurate mixing of small volumes inside the meandering channels of the quartz chip and allows acquisition of thousands of SERS spectra with 785 nm excitation at an integration time of 1 s. Preparation of three batches of three leukemia cell lines demonstrated the experimental reproducibility. The main advantage of a high number of reproducible spectra is to apply statistics for large sample populations with robust classification results. A support vector machine with leave-one-batch-out cross-validation classified SERS spectra with sensitivities, specificities, and accuracies better than 99% to differentiate Jurkat, THP-1, and MONO-MAC-6 leukemia cell lysates. This approach is compared with previous published reports about Raman spectroscopy for leukemia detection, and an outlook is given for transfer to single cells. A quartz chip was designed for SERS at 785 nm excitation. Principal component analysis of SERS spectra clearly separates cell lysates using variations in band intensity ratios.