Angular contact ball bearings are widely used in the field of rotating machinery due to their obvious advantages such as relatively good positioning accuracy, high speed rotating performance and low cost, which have already become the most important transmission components. The heterogeneous thermal deformation caused by the high speed effect of bearing parts will lead to excessive noise and even gluing, which can further significantly reduce the machining accuracy of machine tool. Therefore, it is vital to improve the deformation resisting capability and operational reliability of the whole system. For this purpose, this paper presents a reliability model for computerized numerical control (CNC) lathe spindle by considering thermal effect. A five-degree-of-freedom quasi-static model considering thermal deformation is firstly proposed to calculate contact load and contact angle. Then the transient thermal network method is used to solve the temperature value of multi-node spindle-bearing system, and the validity of the proposed model is verified by experiments at different speeds. Next the modified first-order and second-moment method (FOSM) is used to calculate the reliability and reliability sensitivity of CNC lathe spindle deformation model considering thermal effect. Finally, the constrained nonlinear optimization method for the reliability model is proposed and applied to CNC lathe spindle. The results show that the reliability of the optimized model is significantly improved and the reliability robustness is enhanced.