Cement-mixing method is gaining popularity as a method for stabilizing soft soils in applications ranging from the improvement of foundation properties to mitigation of liquefaction. However, spatial variability in the shear strength of the cement-treated ground introduces uncertainties in estimating the bearing capacity for design. This paper presents a reliability assessment for the bearing capacity of cement-treated ground based on the results of a probabilistic study in which the shear strength of the cement-treated ground is represented as a random field in Monte Carlo simulations of undrained stability for a surface foundation using numerical limit analyses. A statistical interpretation of the bearing capacity is based on Monte-Carlo simulations using the Random Field Numerical Limit Analyses. The results show how the bearing capacity is related to the coefficient of variation and correlation length scale in the shear strength of cement-treated ground. Based on the result, an overdesign factor, percent defective and resistance factor in LRFD for the bearing capacity of cement-treated ground are proposed to obtain a target reliability index.