In September 2024, the US Department of Energy’s Advanced Scientific Computing Research program convened a Workshop on Energy-Efficient Computing for Science to address the critical research challenges and opportunities in this field. The workshop brought together experts from academia, government, and industry to explore innovative approaches to improve energy efficiency across the computing stack over the next two decades. Participants identified five priority research directions (PRDs) that emphasize the need for a holistic approach. These PRDs include (a) codesigning energy-efficient hardware devices and architectures for important workloads; (b) defining the algorithmic foundations of energy-efficient scientific computing; (c) reconceptualizing software ecosystems for energy efficiency; (d) enabling energy-efficient data management for data centers, instruments, and users; and (e) developing integrated, scalable energy measurement and modeling for next-generation computing systems. The workshop’s findings emphasize that achieving significant advancements in energy efficiency requires a multifaceted strategy. This includes innovations in hardware, algorithms, software, and data management as well as crosscutting efforts in modeling, simulation, and benchmarking. In addition, the participants identified multiple enablers for technical progress in this area, such as computing test beds and software maintainability, which are crucial for rapid progress in this area. The research directions outlined in the workshop report aim to guide the development of energy-efficient computing technologies that can support future scientific discoveries and address the growing energy demands of scientific computing and AI.