Decision Making in Energy Sourcing Strategies for Data Centers Under Rapid Energy Demand Growth : A study on risks and trade-offs in the Swedish Market

Sammanfattning

The rapid emergence of Artificial Intelligence technologies on the market has led to accelerated electricity consumption within the data center sector. This increase has placed pressure on national grids and extended timelines for securing electricity connections. Increasing grid queue times and price uncertainty have placed greater pressure on decision makers to find alternative electricity procurement strategies for data centers. This master's thesis explores how the rapid energy demand growth is altering energy sourcing decision making for data centers in Sweden and how these new potential challenges can be handled. This thesis adopts a mixed-methods research approach with both qualitative and quantitative insights. The qualitative data was gathered through structured and semi-structured interviews with a diverse set of professionals as respondents. Quantitatively, the study applied multi-criteria decision-making (MCDM) models like VIKOR (Vise Kriterijumska Optimizacija I Kompromisno Rešenje) and TOPSIS (Technique for Order of Preference by Similarity) to analyze different energy sourcing alternatives based on a specific set of criteria. The MCDM models were applied on both data from a pre-AI-induced electricity demand scenario (2019) and current data, to analyze shifts in decision making priorities. The findings show that risks associated with energy sources for data centers are increasing. Relying strictly on connecting to the grid via spot market pricing is becoming a less viable option. Future regulatory uncertainties connected to risks related to price and power delivery are making other alternatives increasingly attractive. Risk mitigating strategies like Power Purchasing Agreements (PPAs) are becoming more viable due to their ability to reduce price- and regulatory risks. Additionally, the study points to a shift from isolated energy sourcing strategies for data centers towards collaborative approaches with multiple actors. Data centers can strengthen their resilience by focusing on strategies that enhance public acceptance, such as supporting new renewable energy projects or utilizing excess heat for district heating. Strategies with backup power technologies can become increasingly important to reduce peak stress on the grid. Lastly, this thesis shows how MCDM can help decision makers compare and evaluate different energy sourcing strategies to make more informed decisions. This study contributes to research on decision making in the field of energy sourcing for data centers and risk management in energy sourcing, and strategic management during periods of rapid technological disruption. Specifically, the study advances these fields by demonstrating how MCDM methods can be applied to structure and compare energy sourcing alternatives under conditions of high uncertainty. By applying the models across both a pre-AI demand scenario (2019) and a current scenario (2025), the study provides a structured comparison of how decision-making priorities have shifted in response to AI-driven energy growth. Furthermore, the study contributes by examining how social acceptance and societal value creation are becoming structural factors in energy sourcing strategy for data centers in the Swedish market.