Nowadays sustainable underground geothermal energy resources have received special attention thanks for being characterized as clean, zero-carbon footprint, reliable, and free source of renewable energy that can run all year long and around the clock. Barren desert lands, which make up 33% and contribute to almost 30 Million km² of global land surface area, is increasingly seen as supply of green energy but not yet efficiently and globally utilized although it can save up to 70% compared to traditional HVAC systems bills. This paper presents a novel artificial intelligent machine learning and big data algorithm to analyze and control geothermal heat pump system (GHP). In particular, the main objective of this research is to model, design, analyze, control and optimize the performance of desert underground GTH system based on thermodynamics laws and AI machine learning. As a case study, the analysis and design of desert GHP is performed based on the annual weather data collected for Al Ain city in UAE. By selecting a horizontal layout, the design analysis results show that GHP unit needs a 66 m total trench length with a cooling capacity estimated of 12.4 kW, heat pump COP of 2.8 and 1.6 for the system COP with 30.3 L/min water flow rate. Similar results for the heating system are obtained as well. Furthermore, financial calculations show the GHP system is very economic and competitive comparing with the traditional cooling/heating systems. It is figured out that the annual cost of the GHP system costs around $1676 compared with $7992 if air-cooled chiller and boiler are used. To maintain the geothermal system for one life cycle (usually 20 years) it needs to spend only $14,659 compared with $109,944 in case HVAC system is utilized. The overall life cycle cost in case of the desert GHP system does not exceed (45%) of the traditional HVAC system ($81,881 compared to $181,974). One of the direct applications is use this proposed desert GHP to cool the roof water tank for domestic and personal usage. Furthermore, artificial intelligent and big data machine learning is executed to analyze the weather conditions related to the GHP performance based on huge number of thermal observations recorded for the years 2015-2018. Moreover, the mean switch-off control hours of the GHP is examined by developing a supervised learning predictive model. For the purpose of validation a four ton Bosch GHP unit is selected as a benchmark. Switch-off control hours per month for the entire geothermal data set are demonstrated by using a linear regression model that help to guide the controller to switch-on/switch-off the system without having the need for the real data measurement. One primary outcome obtained is the ability to optimize the GHP performance, save primary input energy and operation periods. Furthermore, the results interprets that almost one third of the year is in a switched-off saving mode (33%), compared to 67% in switch-on mode. This smart big data control will lead to a life-cycle saving of $27,020. This AI saving strategy is found to be competitive and leading compared to other schemes. It is worthy to recommend linking GHP controller with real-time radar or weather station that will fed the system with real data conditions which would lead to improving its performance and dispense costly measuring sensors.