Abstract: To meet the technical requirements for real-time and accurate acquisition of rock mechanical parameters during deep resource drilling, this study proposes a cross-scale method for predicting mechanical properties using drilling cuttings, combining atomic force microscopy (AFM) and laser Raman spectroscopy. Using granodiorite from the Gonghe Basin as the research object, a systematic cross-scale mechanical analysis was conducted from the microscopic to macroscopic levels. Laser Raman spectroscopy was employed to perform surface scanning tests on rock thin sections, and together with image processing algorithms, enabled rapid mineral identification and quantitative characterization of spatial distribution. Atomic force microscopy testing provided the elastic modulus distribution within different mineral grains and at their boundaries, revealing significant mechanical weakening effects at mineral interfaces. Based on microscopic test results and the particle flow discrete element method, mineral-scale numerical models of granodiorite were constructed. Uniaxial compression and direct tensile loading simulations were performed, successfully predicting the macroscopic mechanical properties of the rock, with simulated results showing good agreement with experimental values. Microscopic failure analysis indicated that tensile cracks within minerals dominate under compressive conditions, while penetrating cracks perpendicular to the loading direction primarily form under tensile conditions. The randomness of mineral spatial distribution showed relatively limited influence on the macroscopic mechanical properties of the rock. Compared with traditional testing methods, this approach eliminates the need for coring, offers advantages of shorter testing cycles and enhanced timeliness, provides a new technical pathway for rapid acquisition of rock mechanical parameters, and is expected to facilitate the transition of drilling evaluation methods from centimeter-scale cores to millimeter-scale cuttings.