STUDY OF WASTE HEAT RECOVERY SYSTEMS FROM DATA CENTERS FOR INDUSTRIAL APPLICATIONS

The rapid growth of digital services and generative artificial intelligence is driving an unprecedented rise in data center energy demand, estimated at approximately 485 TWh globally in 2025, with IEA projections indicating a possible doubling by 2030. Nearly all the electricity absorbed by IT equipment is converted into heat through the Joule effect, and in most cases this heat is rejected to the environment, with significant energy and environmental consequences, including the so-called data heat island effect. This thesis analyses the main technologies for recovering and valorising this thermal resource, namely Organic Rankine Cycles (ORC), heat pumps and Carnot batteries, describing their operating principles and configurations and examining representative case studies from the literature. The analysis shows that all three technologies are technically viable, and that the optimal choice depends on the end use of the recovered energy. Heat pumps are the most suitable option when heat itself is needed, for district heating, buildings, industry or agriculture. ORC systems are preferable when the goal is direct electricity generation, although with modest conversion efficiencies, below 10%. Carnot batteries are best suited to cases where energy must be stored and returned as electricity during peak-price periods, despite exhibiting a round-trip efficiency of around 60%.