To answer the growing demand for data computing, data storage and data transmission, the current trend is to construct centralized hyperscale data centers, whose power usage effectiveness (PUE) has substantially improved in recent years. Still, the operation of hyperscale data centers has economic and environmental impacts on society, especially if their residual heat is not reused. This study aims to present the economic and environmental rationales of decentralized computing capacities integrated into the building stock. The proposed solution simultaneously tackles the mitigation of greenhouse gas emissions from the residential sector and the ever-increasing global data demand, by exploiting the synergy between domestic thermal needs and digital services, while integrating PV production. The originality of this work resides in its holistic approach and multi-services (data, heat and electricity) dimension. A Mixed Integer Linear Programming (MILP) energy modelling optimization framework is developed to assess the economic and environmental performance of centralized hyperscale infrastructures versus decentralized building-integrated computing units with heat recovery. Results for the decentralized solution underline the reduction of environmental impact (-16.3%) and global cost (-27.7%) for the benefits of the whole society. As part of an integrated system, these decentralized computing units foster PV penetration, avoid the oversizing of a heat pump, and achieve high self-consumption (96%), while preventing the installation of a battery through coordination between thermal storage, data services and solar availability. This efficient solution offering an unmatched energy reuse effectiveness (ERE) shows the whole potential of the digitalization of energy systems, is a key contribution to building energy systems design, and reveals a promising approach to mitigate the increasing energy consumption of the ICT sector.