Cancer progression is a complex, multi-stage development process characterized by dynamic changes at the molecular level. Understanding these changes may provide new insights into tumorigenesis and potential therapeutic targets. This study focuses on the evolutionary transcriptomics of cancer, specifically analyzing the Transcriptome Age Index (TAI) across different pathological stages. By examining various cancers at four distinct pathological stages, we identify a significant «hourglass» pattern in TAI indices of ductal carcinoma of the breast, bladder carcinoma, and liver carcinoma, suggesting a conserved evolutionary trajectory during tumor development. The results reveal that early and late stages of these cancers exhibit higher TAI values, indicative of more novel gene expression, while intermediate stages show a dip in TAI, reflecting a more ancient evolutionary origin of expressed genes. This «hourglass» pattern underscores the evolutionary constraints and innovations at play during tumor progression. Our findings contribute to the growing body of evidence that evolutionary principles are deeply embedded in cancer biology, offering new perspectives on the dynamics of gene expression in tumors.