In vivo, molecular-level investigation of cytokinesis, the climax of the cell cycle, not only deepens our understanding of how life continues, but it will also open up new possibilities of diagnosis/prognosis of cancer cells. Although fluorescence-based methods have been widely employed to address this challenge, they require a fluorophore to be designed for a specific known biomolecule and introduced into the cell. Here, we present a label-free spectral imaging approach based on multivariate curve resolution analysis of Raman hyperspectral data that enables exploratory untargeted studies of mammalian cell cytokinesis. We derived intrinsic vibrational spectra and intracellular distributions of major biomolecular components (lipids and proteins) in dividing and nondividing human colon cancer cells. In addition, we discovered an unusual autofluorescent lipid component that appears predominantly in the vicinity of the cleavage furrow during cytokinesis. This autofluorescence signal could be utilized as an endogenous probe for monitoring and visualizing cytokinesis in vivo.