The High-Luminosity LHC (HL-LHC) era will bring a substantial increase in the computing demands for the Worldwide LHC Computing Grid (WLCG), posing important challenges in terms of costs, energy consumption, and environmental sustainability. In this context, WLCG has established a Sustainability Forum to coordinate community-wide efforts aimed at reducing the environmental footprint of HEP computing. This talk presents an overview of WLCG's findings, constraints, and their sustainability strategy, highlighting how advancements in software and hardware acceleration can translate into important energy savings. At the same time, power-aware benchmarking through the HEPScore23 suite provides insights into the impact of hardware choices and configurations. As an illustrativea main example how a reduction can be achieved, we discuss recent developments in Monte Carlo event generation, namely MadGraph (using a plugin for hardware acceleration) and PEPPER, where GPU acceleration and CPU vectorization have demonstrated considerable performance gains and measurable reductions of energy consumption per event. Furthermore, we discuss other activitiesinitiatives, like such as GPU acceleration of detector simulation with AdePT and Celeritas, and machine-learning-based fast simulation with FlashSim. These efforts show that targeted software improvements, combined with the adoption of heterogeneous architectures, represent a viable path toward more sustainable computing for the HL-LHC era.