The detection of kidney cancers at an early stage is critical for diagnosis and therapy. Surface-enhanced Raman scattering (SERS) is investigated for early detection of cancer cases from biopsy samples. The colloidal silver nanoparticles as the SERS-active nanostructures are directly mixed with homogenized tissue samples. The SERS spectra from the normal and abnormal tissue samples collected from 40 cancer patients, 28 of them at T1 stage and 12 of them at T2–T3 stages, are analyzed using principal component analysis combined linear discriminant analysis with leave-one-out cross-validation method. It is found that the diagnosis sensitivity, specificity, and total accuracy of the approach can be as high as 100%. The results suggest that SERS can be used as a potential technique for the identification of the different tumor stages.