The statistics of the temperature and thermal stress are quantitatively obtained by Monte Carlo simulation method for functionally graded material plates (FGM plates) with uncertainties in the thermal conductivity and coefficient of linear thermal expanssion and with arbitrary nonhomogeneous thermal and mechanical properties through the thickness of plate. The stochastic temperature and thermal stress fields are analyzed by approximating the FGM plates as multilayered plates with random thermal conductivity and coefficient of linear thermal expanssion in each layer. Autocorrelation coefficients of the random properties and crosscorrelation coefficients between the random properties are expressed in exponential function forms as homogeneous Markov random field of discrete coordinate. Numerical calculations are carried out for FGM plates composed of PSZ and SUS304, which have two different variability of random properties. The effects of the step-wise change in material composition, boundary condition and correlation coefficients on the standard deviations of temperature and thermal stress are discussed.