As the last barrier of nuclear materials, the concrete containment is often in a rapidly changing extreme condition in the event of an accident. This study uses artificial intelligence (AI) techniques to establish a correlation between the position of the containment and stress-strain values under such extreme working conditions, aiming for efficient and accurate prediction of mechanical properties. Finite element (FE) software is utilized to analyze the containment structure, generating batches of prestressed concrete models under various internal pressure conditions. The internal pressure value and three-dimensional coordinates of the containment are considered as input values in artificial intelligence algorithms, while the maximum principal stress or equivalent plastic strain as output values. This work shows that AI models can offer significant advantages in terms of time efficiency by replacing laborious finite element simulation modeling, analysis, and post-processing procedures while enabling predictions regarding containment structural stresses under extreme conditions. Furthermore, these trained models possess remarkable generalization capabilities allowing them to predict stress and strain for a wider range of internal pressure scenarios.