[Objective] The crack initiation mechanism of saturated frozen sandstone under compression remains insufficiently understood, particularly regarding the quantitative relationship between initiation stress and peak stress under varying confining pressures, failure mode transitions, and the influence of micro-crack orientation on fracture resistance. This study aims to: quantify initiation stress via the volumetric strain inflection method; establish its relations with confining pressure and peak stress; characterize failure mode evolution; reveal the anisotropic effect of crack orientation; and interpret the initiation mechanism from an energy dissipation perspective. [Methods] Cylindrical saturated sandstone specimens (50 mm×100 mm) were frozen at -10 ℃ for 48 hours. Triaxial compression tests were conducted under confining pressures of 0, 2, 4, and 6 MPa at 0.1 mm/min. Axial and circumferential strains were continuously monitored. Crack initiation stress was determined from the inflection point on the crack volumetric strain curve. At least three parallel specimens were tested at each pressure to ensure reproducibility. [Results] (1) Both peak and initiation stresses increase with confining pressure. Initiation stress ranges 25%-35% of peak stress across all confining pressures, while peak stress shows a greater growth rate.(2) Strain evolution differs by stress state. Under uniaxial compression, circumferential strain stabilizes and volumetric strain dilates while axial strain continues increasing. Under triaxial compression, both strains continue increasing with different offsets, reflecting lateral constraint.(3) Failure modes transition with confining pressure. Uniaxial compression produces axial splitting (mode I tensile) with vertical cracks and flat surfaces. Triaxial compression shifts to shear-dominated failure with inclined shear bands and stepped surfaces.(4) Micro-crack orientation significantly affects fracture resistance. When the long axis of pre-existing micro-cracks is oriented at 30° to the loading direction, crack initiation occurs most readily. This angle matches the internal friction angle of the sandstone matrix, indicating that stress concentration is maximized when cracks align with the potential shear plane, promoting wing crack nucleation. [Conclusions] Based on the experimental findings, the following conclusions are drawn. (1) The crack initiation stress of saturated frozen sandstone, as determined by the volumetric strain inflection method, represents a reliable and practical indicator of the damage threshold. The finding that initiation stress consistently comprises 25%-35% of peak stress across different confining pressures provides a quantitative basis for engineering design, enabling conservative estimates of the safe loading range in frozen rock masses.(2) The anisotropic control of fracture resistance by micro-crack orientation is a critical factor in assessing the stability of frozen rock engineering structures. The identification of 30° as the most unfavorable angle, consistent with the internal friction angle of the rock matrix, implies that geological surveys should prioritize the characterization of micro-crack fabric when evaluating potential failure planes in frozen sandstone formations. Ignoring this anisotropy may lead to non-conservative stability predictions.(3) The initiation of cracking in saturated frozen sandstone is fundamentally an energy-driven damage evolution process rather than a simple stress-induced failure. Strain energy progressively accumulates at the grain-ice crystal interfaces, primarily in the form of elastic strain energy stored in the frozen bonds, until reaching a critical threshold at the volumetric strain inflection point. Beyond this threshold, the stored energy is dissipated through two concurrent micromechanisms: tensile micro-cracking and shear micro-cracking. The competition and synergy between these two mechanisms control the overall macroscopic failure behavior. This energy-based interpretation provides a unified framework for understanding the coupled effects of stress state, micro-crack fabric, and ice bonding on the mechanical behavior of frozen geomaterials.(4) These findings have direct practical implications for the design and stability assessment of cold-region geotechnical infrastructure, including tunnels, slopes, and foundations in permafrost areas. Future work may extend the present methodology to variable temperatures, different degrees of saturation, and cyclic loading conditions to more comprehensively capture the time-dependent and environmental-dependent cracking behavior of frozen rocks.