Abstract-Energy-efficient network design has recently become a very important topic because of the energy cost increases in service providers' infrastructures. This is of particular importance in access networks because of the growing demand for digital traffic by end users. Here we address the challenge of reducing the energy consumption of fiber-wireless (FiWi) access networks, which use both optical and radio frequency technologies to provide high bandwidth and ubiquity for end-user applications, while keeping delay under a threshold. Our goal is to find optimal sleep mode schedulings that allow energy consumption to be reduced while keeping packet delay acceptable. For this purpose a mathematical formalization and an algorithm are developed. The results show that the proposed approach is able to reduce the average packet delay, with negligible energy cost increases, in many scenarios, besides being computationally efficient and scalable. The proposed approach may, therefore, serve as a basis for planning and design of quality-of-service-aware energy-efficient FiWi access networks.