Background: Cancer subtypes classified by DNA microarray data have shown excellent abilities in predicting patient prognosis; in particular, consensus molecular subtypes (CMSs) are regarded as the most robust classification system with clear biological interpretability. Recently, we performed unsupervised clustering analysis using a public database and selected the expression of 55 gene to construct a discriminant model with the aim of classifying colon cancer (CC) into three subtypes: “microsatellite instability (MSI)-like”, “chromosomal instability (CIN)-like”, and “stromal”; the recurrence rates of these subtypes were shown to be different (p = 0.001). In this study, we conducted a retrospective, multi-institutional study to validate the quality of a novel 55-gene classifier (55GC) for Stage II CC and compared 55GC subtypes with the CMS categories. Methods: We collected formalin-fixed, paraffin-embedded cancer specimens from 232 patients with Stage II CC who underwent curative surgery (R0) without adjuvant chemotherapy at 10 institutions between 2009 and 2012. Tissue sections were prepared from each specimen and subjected to DNA microarray measurement. Results: Using the 55GC, patients were classified as having the MSI-like subtype in 27%, CIN-like in 41%, and stromal in 32% of the cases. The 5-year recurrence-free survival (RFS) rate of patients with the MSI-like, CIN-like, and stromal subtype cancers was 88.5%, 83.3%, and 71.2%, respectively (stromal vs. others: p = 0.0049). Multivariate analysis by Cox's proportional hazard model revealed that stromal subtype (hazard ratio, 2.3; p = 0.0063), pT4, and the number of lymph nodes examined ( < 12) were independent poor prognostic factors. The overall concordance rate between 55GC and CMS was 72%: 29% (18/62) of the MSI-like subtype, 79% (75/95) of the CIN-like subtype, and 98% (74/75) of the stromal subtype were judged as CMS1, CMS2/3 and CMS4, respectively. The 5-year RFS rate of patients with CMS1, CMS2/3 and CMS4 was 100%, 86.3%, and 73.0%, respectively (CMS4 vs. others: p = 0.005). Conclusions: 55GC is a useful and reproducible grading system for recurrence risk stratification of Stage II CC.