Climate change poses significant challenges to infrastructure in cold regions, particularly in permafrost. Predicting ground surface temperature is essential for understanding the impacts of climate change and the planning, design, and management of climate-resilient infrastructure in these regions. However, existing methodologies face gaps and challenges, such as surface energy balance complexity, non-stationary correlations between climate variables, and the scarcity of historical ground temperature measurements in remote areas. This thesis develops artificial intelligence (AI) methodologies to predict ground surface temperature and assess its impacts on infrastructure in cold and permafrost regions. An AI-powered methodology using machine learning models was developed to forecast long-term ground surface temperature from other climate variables. The predictions were used to assess the performance and sustainability of shallow geothermal systems in cold climates. A multi-variable evaluation of ERA5 and ERA5-Land datasets was then conducted in order to utilize reanalysis data for AI model training. The methodology was further advanced using long short-term memory (LSTM) networks and multiple predictor variables from ERA5-Land data, including air temperature, irradiance, wind speed, and snow depth. A framework was then developed to assess climate-driven permafrost thaw threat using ground surface temperature and ground ice distribution data. It was applied to assess threats to three major northern Canadian land transportation infrastructure, including the Hudson Bay Railway, the Mackenzie Northern Railway, and the Inuvik–Tuktoyaktuk Highway. A comparative analysis of surface boundary condition methods for thermal geotechnical analysis was also conducted by evaluating the AI-powered approach, the n-factors, and the surface energy balance heat flux. The AI-based approach outperformed traditional methods. The feasibility of climate-driven geotechnical simulations using an AI-predicted surface boundary condition was demonstrated through a case study in Salluit, Nunavik. Finite element models indicated significant changes in the active layer thickness and inferred supra-permafrost talik under future climate scenarios, highlighting substantial risks to infrastructure due to climate change.